By Kent R. Kroeger (Source: NuQum.com, May 5, 2020)

Note: In this article I use the terms ‘lockdown’ and ‘stay-at-home’ orders interchangeably.

There is a lot of partisan noise in the news coverage of the coronavirus (COVID-19) pandemic.

“This isn’t about science, this is about control,” Fox News host Laura Ingraham recently told her audience as she questioned the motives of those who believe the U.S. must continue to lockdown until the COVID-19 pandemic is more under control.

Meanwhile, on MSNBC, Chris Hayes was calling the hosts at Fox News “coronavirus truthers” for giving airtime to two California urgent care doctors who believe we no longer need statewide lockdowns, suggesting the cable new network’s coordinated push to “get Americans back to work” was not based on science but on ideology and blind greed.

The “lockdown” is the ideological pivot point for the current U.S. political discourse on the COVID-19 pandemic.

In dispute are three fundamental issues:

(1) What is the actual morbidity (infection) rate of the coronavirus (SARS-CoV-2)?

(2) What is the actual mortality rate of the COVID-19?

(3) How effective have the statewide locksdowns been in containing and suppressing COVID-19?

Multiple large-scale studies based on probability samples will be needed to answer these questions definitively. Suggestions that the science is already established is absurd.

Only four days ago, New York Governor Andrew Cuomo announced results from a Staten Island, New York SARS-CoV-2 antibody study that found 20 percent of residents had been exposed to the virus. For the State of New York, overall, Cuomo said 12 percent had tested positive for the coronavirus.

If the New York results can be extrapolated to the entire U.S. population (it can’t, but for arguments sake), that would mean between 39 and 65 million Americans have had the virus. As of 3 May, around 70,000 Americans have died from COVID-19, meaning a rough estimate of the mortality rate of COVID-19 is somewhere between 0.1 and 0.2 percent — a range that covers the mortality rate for an average flu season and a really, really bad flu season.

Fox News and those two California doctors should have received an apology from Chris Hayes by now.

But the reality is, this New York State and Staten Island studies are just the beginning of the research that needs to be conducted over the next year or two before we know the actual morbidity and mortality rates of COVID-19.

As we are still in the middle of this health crisis, any definitive proclamations one way or the other are generally dubious…including mine.

Have Statewide Lockdowns Helped?

It is not too early to start assessing the effectiveness of statewide “stay-at-home” orders and economic lockdowns, as whether will end them in a same and timely fashion will depend on some quantitative level of knowledge about their utility.

A superficial look at the data suggests the lockdowns are not helping much (Note: This is not my opinion on lockdowns. I am doing this data walk in order to move towards a more evidence-based opinion)

While the states hit the hardest by the COVID-19 pandemic (on a per capita basis) are mostly Democrat-dominated states that decided relatively early in the crisis to issue statewide lockdown (“Stay-at-Home”) orders, states that adopted less stringent lockdown policies (Arkansas, Iowa, South Dakota, Nebraska, North Dakota, Utah and Wyoming) appear to be doing better (see Figure 1).

Figure 1. The relative number of COVID-19 cases and deaths by U.S. State (through 3 May)

Of the 10 states (plus District of Columbia) with the highest relative number of COVID-19 cases, only two were won by Donald Trump in 2016 (Louisiana and Michigan). Conversely, among the 10 states with the lowest relative number of cases, seven were won by Trump (Montana, Alaska, West Virginia, Wyoming, Oklahoma, Texas and Kentucky).

States won by Trump in 2016 average 1,885 COVID-19 cases per 1 million people, whereas states won by Hillary Clinton average 4,818 per 1 million people. Though this difference is significant using a two-sample difference of means t-test (t-statistic = 3.47, p = .001, std. error = 846), in no way can we conclude that Trump (“red”) states are doing better than Clinton (“blue”) states.

Furthermore, the real analytic question is about the effectiveness of “Stay-at-Home” orders, which some red states did implement. In fact, the difference in the relative number of cases between lockdown (“Stay-at-Home”) states and non-lockdown states is not statistically significant; in part, because we have so few instances of non-lockdown states (n = 7).

Also complicating our ability to discern the effectiveness of lockdowns is the variation in lockdown policies across states. Indeed, even a few non-lockdown states (such as Iowa) have implemented elements of a lockdown (such as social distancing recommendations, closing bars and restaurants and forbidding large group gatherings). Perhaps those policies, not a full lockdown, are sufficient to contain the spread of COVID-19?

We must also consider that the state-level differences in Figure 1 could be driven by factors unrelated to lockdowns or other strict suppression policies (e.g., travel restrictions within and between states). For example, previously I’ve argued population density is the dominant variable in explaining state-level differences in the relative number of COVID-19 cases, not mitigation and suppression policies. Trump states are doing better in containing the COVID-19 spread in part due to those states being sparsely populated. At a minimum, we need to control for that before declaring which states are doing a better job than others.

If we further divide lockdown states into two categories — those that locked down within 20 days of their first confirmed COVID-19 case and those that took more than 20 days — we might get a clearer picture of the impact of lockdowns.

Figure 2 shows the three lockdown categories across key measures of the coronavirus. With the exception of testing, there is a linear relationship between the three lockdown categories: Early lockdown states are the highest, late lockdown states are in the middle, and non-lockdown states are lowest on key measures such as population density, relative number of cases and deaths, and changes in cases and deaths since 1 April.

Figure 2: Lockdown Categories by Key Measures (as of 3 May)

Early lockdown states are much more densely populated (690 people per sq. mile), have done more testing (25,290 per 1 million people), have more confirmed cases (4,012 per 1 million people), have more COVID-19-related fatalities (230 per 1 million people), and have experienced larger changes in cases (3,306 per 1 million) and fatalities (212 per 1 million) since 1 April.

If Figures 1 and 2 were all you knew about state lockdowns, it would be hard to declare them successful. In fact, similar “topline” data is repeatedly used in the conservative media to reinforce the narrative that the U.S. should return to work sooner rather than later.

Are they wrong?

I believe as a blanket statement — Lockdowns are not working or necessary any more— they are partially wrong. However, to support my opinion, we need to make sure our state comparisons are made on a level playing field.

Can we compare COVID-19 case and death numbers between states?

The definitional comparability problem between states in measuring COVID-19 cases and deaths is highlighted by the Centers for Disease Control (CDC) in its own tabulations and research on the virus (SARS-CoV-2) and its associated disease (COVID-19). In the CDC’s latest report on COVID-19, here is the language they use to explain why their numbers may be different from state-reported numbers:

Provisional death counts in this (CDC) report will not match counts in other sources, such as media reports or numbers from county health departments. Death data, once received and processed by National Center for Health Statistics (NCHS), are tabulated by the state or jurisdiction in which the death occurred. Death counts are not tabulated by the decedent’s state of residence. COVID-19 deaths may also be classified or defined differently in various reporting and surveillance systems. Death counts in this report include laboratory confirmed COVID-19 deaths and clinically confirmed COVID-19 deaths. This includes deaths where COVID-19 is listed as a “presumed” or “probable” cause. Some local and state health departments only report laboratory-confirmed COVID-19 deaths. This may partly account for differences between NCHS reported death counts and death counts reported in other sources.

My state of New Jersey, for example, only reports COVID-19 deaths that are test-confirmed, except in the case of nursing home deaths where “probable” cases can also be classified as a COVID-19 related death. This may account for why nursing home deaths account for nearly half of New Jersey COVID-19 deaths. Unfortunately, the New Jersey fatality numbers in nursing homes must therefore be considered problematic even if, in the words of Jon Dolan, president of the Health Care Association of New Jersey, the data are “directionally accurate.”

Yowza! Directionally accurate?!

Worse yet, these case and fatality definitions can vary depending on the state where the victim died, according to the CDC.

This is a serious problem if the goal is to compare state-level mitigation and suppression policies. Without detailed case-level data on COVID-19, researchers are at a distinct disadvantage.

Adding to this problem is an even bigger problem, in my opinion. There is significant variation in how diligently states are trying to measure the spread of the coronavirus within their state. As of 3 May, Rhode Island has conducted 67,896 per 1 million people, and New York 50,680 per 1 million. Contrast those numbers to Arizona and South Carolina where, respectively, only 11,144 and 12,467 tests per 1 million have been conducted so far.

Does this mean New York or Rhode Island’s estimate of the morbidity rates (percent of residents that have had the virus) in their respective states are more accurate than Arizona and South Carolina? Not necessarily. It will depend on a number of factors related to the quality of a state’s effort to assess the spread of the coronavirus (see Figure 3), including: (1) COVID-19 test availability, (2) training in using the test, (3) test accuracy, and (4) and selection bias (perhaps the most serious problem if the goal is to generate an unbiased estimate of a state’s morbidity rate). We must also recognize that all of these factors are moving together in time. What was true on Tuesday may not be as true on Friday.

Figure 3: A Model for Assessing the Quality of a State’s COVID-19 Measurement Effort

If a state is only testing health care workers and citizens with severe symptoms (as New Jersey was doing initially), their testing will miss a significant percentage of people who are asymptomatic but still contagious. That is textbook selection bias. Only some form of probability sampling will reliably measure the prevalence of this group or the overall morbidity rate in a geographic location. Some states (hopefully most) are doing that already (New York is, for example). And I am certainly not the first person to call for more random sampling in measuring the COVID-19 pandemic (others doing the same are here, here and here).

Ironically, its the supposedly anti-science Republicans in North Carolina that have pushed for probability sample testing, not the Democrats. Why? The Democrats fear random testing will reveal their fundamental rational for imposing state-level lockdowns — a high COVID-19 mortality rate relative to the common flu — likely won’t be substantiated by the science.

The problem too is that we have a lazy, politically-biased national journalism corps in this country that would rather rely on the deceptive case numbers and case fatality rates pumped out by Johns Hopkins University (JHU-CSSE) and the WHO than insist on scientifically accurate numbers, such as the estimates recently released by a Stanford University study in Santa Clara County, California, showing a significantly higher COVID-19 prevalence rate than the official numbers released by the county and state governments.

“After adjusting for population and test performance characteristics, we estimate that the seroprevalence of antibodies to SARS-CoV-2 in Santa Clara County is between 2.49% and 4.16%, with uncertainty bounds ranging from 1.80% (lower uncertainty bound of the lowest estimate), up to 5.70% (upper uncertainty bound of the highest estimate),” the study said. Those estimates translate to between 48,000 and 81,000 Santa Clara County residents having had the coronavirus (SARS-CoV-2).

At the time of the study’s field closure, April 1st, the State of California and Santa Clara County were reporting a case prevalence of just 956 in the county. The implication of this difference from the Stanford results is profound: If the Stanford study is correct, the mortality rate for COVID-19 is not nearly as high as being implied when the news media reports the Johns Hopkins University case and fatality numbers. If one of the fundamental arguments behind imposing statewide lockdowns is the purportedly high mortality rate for COVID-19 — as high as six percent if using only confirmed cases (as reported by JHU-CSSE) in the denominator — any evidence that pushes that number closer to 0.1 percent (the estimated mortality rate for the common flu) is heresy.

When two urgent care doctors in California suggested as much, YouTube attempted to remove their televised news conference from its social media platform. Facebook and Twitter, at least so far, have resisted the same action.

We don’t need to use the numbers offered by the two California doctors to argue the media-reported COVID-19 mortality rates may be seriously inflated — we have the Stanford study. But is the Stanford University necessarily right? Of course not, but at least that study’s team documented their methodology, giving readers (and the news media) something to critique.

An even larger scientific study in China found a higher fatality rate for COVID-19 than the Stanford study —2.3 percent overall and as high as 14.8 percent for people aged 80 and older. Across other age groups, the fatality rate was found to be as follows:

• Aged 70 to 79 = 8.0%
• Aged 60 to 69 = 3.6%
• Aged 50 to 59 = 1.3%
• Aged 40 to 49 = 0.4%
• Aged 0 to 39 = 0.2%

Let me now reveal my bias regarding statewide lockdown policies: If the Stanford study is correct and the COVID-19 fatality rate is similar to the common flu, it is time to end the statewide lockdowns (while maintaining strict social distancing recommendations). However, if the Chinese study is more accurate, as a man in his mid-50s, a 1.3 percent chance of dying from COVID-19 is too hot for my oatmeal — I’m staying home until there is a vaccine or reliable treatment (and anyone my age or older should do the same).

Which is why the national news media’s attempt to suppress the important scientific debate going on right now about the mortality rate of COVID-19 is reckless and counterproductive.

We may be huddling in our basements for no reason and to little effect? Is the damage being done to the world economy worth being so overly cautious? I would say, ‘No,’ particularly if people are losing their health care coverage or being saddled with large medical bills due to the pandemic.

Do you want to know a way to kill millions of Americans even faster than COVID-19? Take away their health care coverage.

Enough news media bashing. It has become too easy.

The point of this article is to make a preliminary assessment of the effectiveness of statewide lockdowns in controlling the spread and lethality of COVID-19.

Here is my take on the data (as of 3 May 2020)…

After controlling for population density (and other factors), what can we say about statewide lockdowns to control COVID-19?

My answer is not a satisfying one: it is not clear the states with lockdowns are doing better containing the effects of the coronavirus than states without full lockdowns.

One caveat to this finding is important: the COVID-19 pandemic is not over. Many non-lockdown states like Iowa and Nebraska still have not experienced a decisive peak in new daily cases. But also, keep in mind, these state-reported cases and deaths numbers are of dubious quality (see discussion above). The fact that Iowa Governor Kim Reynolds found the Utah-based company to handle Iowa’s COVID-19 testing program through D-list Hollywood celebrity and Iowa-native Ashton Kutcher should tell you all you need to know about how seriously I take Iowa’s measurement program.

[Sidebar: You would think Iowans would have learned their lessons from the Iowa Caucuses vote counting disaster last February when the Iowa Democratic Party chose another Utah-based tech company to count the votes. Are these people morons? What happened to Iowa’s once premiere public education system? That’s a debate for another article — and a heads-up, my conclusions are not kind to Iowa’s political leadership.]

Another caveat is that this is an aggregate (state-level) data analysis. It does not inform us directly about individual-level relationships. So, when you read below that the relative number of tests, indirectly, has a positive relationship to the relative number of COVID-19 deaths, this does not mean states should stop testing to reduce the number of deaths! The result means in the aggregate that states with the most relative testing are also experiencing higher relative fatality rates for COVID-19, all else equal. Correlation is not causation.

A Path Model for COVID-19 Cases and Deaths for U.S. States & D.C.

Using the JHU-CSSE data at the state-level through 3 May (see, I’m no better than those lazy journalists), I estimated a linear path model for explaining state-level COVID-19 confirmed cases and deaths.

To mitigate the noise and bias inherent in the JHU-CSSE fatality numbers, I calculate the fatality incidence as a function of the state’s total population (i.e., deaths per 1 million people). That is not the mortality rate, but its denominator is a more reliable government statistic than the confirmed number of cases.

The variables tested for their relationship to the relative number of COVID-19 deaths (natural log transformed) included:

• A state’s population density (log transformed) (Variable name: LN_P)
• The cumulative number of tests conducted by the state per 1 million people (natural log transformed) (Variable name: LN_T)
• The number of confirmed COVID-19 cases per 1M people (natural log transformed) (Variable name: LN_C)
• Whether a state imposed travel restrictions between and/or within their state (1 = Yes, 0 = No) (Variable name: TRA)
• Whether a state imposed a statewide (“stay-at-home”) lockdown policy (1 = Yes, 0 = No) (Variable name: STA)
• Governor’s political party (1 = Democrat, 0 = Republican…not a value judgment on my part) (Variable name: PAR)

I also tested variables such as the state’s percentage of vote for Trump in 2016 (sorry, but this confirms to me that the state differences seen in Figure 1 above are not substantively meaningful), GDP per capita, number of staffed beds per 1 million people, percent of state’s economy related to China, and percent of population aged 65 and older. None proved statistically significant.

And here are the results…

First, with only seven states not imposing a full statewide lockdown, it will require sizable predicted differences in relative confirmed case and death rates for the lockdown variable’s two levels to be statistically significant, regardless of whether the model is estimated in a Bayesian or Frequentist context.

The linear path model in Figure 4 was estimated using JASP, a free statistical analysis package (with Bayesian estimation for most analytic options) made available by the University of Amsterdam. The model’s diagnostic output, including standard errors, is in the Appendix below.

Figure 4: Path model standardized parameter estimates and residual covariances (data through 3 May 2020)

Interpretation of Results

Lockdowns versus No Lockdowns

Let us plunge right into the variable of most interest: statewide lockdowns (variable: STA). Its direct relationship to the relative number of COVID-19 deaths is positive and significant (β = 0.90, p < .001), while its direct relationship to the relative number of confirmed cases is negative and significant at an α < .10 level (β = -0.48, p = 0.088).

Have statewide lockdowns led to more COVID-19 deaths? No, not likely. There are anecdotal cases where some victims were isolated in their homes and did not get the urgent care needed to save the lives. One of inquiry in this regard is an increased number of younger and middle-aged victims who died from strokes related to COVID-19, some dying in their homes before they could get needed care. In theory, a statewide lockdown could increase those types of fatalities, but not on the level implied by the model here. But, again, this is an aggregate analysis and we cannot make individual-level inferences.

Instead, my interpretation is that unmeasured factors related to both lockdowns and COVID-19 deaths are at play. Although it is possible that states with lockdowns are inflating their COVID-19 numbers and/or non-lockdown states are under-reporting, I think it is far more likely that some unmeasured variable (my guess is ‘time’) is leading to much higher relative fatality rates in lockdown states compared to non-lockdown states. As all of the non-lockdown states are in the interior of the country, they have been among the last to experience their first cases and deaths compared to the coastal lockdown states.

Still, we cannot rule out the possibility that some states are systematically over-reporting or under-reporting their COVID-19 numbers. I offer the previous discussion about New Jersey including “probable” COVID-19 cases in their nursing home fatality counts as a possible avenue for this type of systemic measurement bias (fraud?). I also have anecdotal evidence from talking to friends back in Midwest about some of their family, friends, and neighbors “toughing it out” and not getting tested despite showing COVID-19-type symptoms. That is a Northern European cultural trait with which I have personal experience.

The more interesting result in our path model is that lockdown states, all else equal, are experiencing lower confirmed case rates per 1 million people than non-lockdown states. While lockdown variable is only marginally significant, that could be for a number of issues unrelated to the actual importance of lockdown policies suppressing COVID-19 morbidity rates.

The confirmed COVID-19 cases model certainly has specification error, though the model residuals cases appear random and the model fit at 69 percent is not too bad (see Appendix).

Over time, the observed differences in relative COVID-19 cases have narrowed between lockdown versus non-lockdown states. When this pandemic is over, we may in fact see much higher relative numbers in the non-lockdown states than we see now. That may, in turn, lead to a stronger (negative) statistical relationship between the lockdown variable and COVID-19 case rates.

Other Factors

When looking at total effects (direct + indirect effects), a state’s population density (+ relationship), relative number of administered tests (+), the presence of travel restriction policies at some point during the pandemic (-), and the presence of a Democratic governor are stronger correlates with the relative number of COVID-19 fatalities (+) than is a state’s lockdown policy. As I’ve said, that could change over time as the pandemic progresses into later stages — the question is: How much?

Residual Analysis

One of the benefits of estimating a linear path model is that we can look at the residuals for each state to see which ones are doing better than expected on COVID-19 case and fatality rates and, of course, which ones are doing worse.

According to my analysis, these states highlighted in green in Figure 5 are experiencing a lower rate of COVID-19 cases (per 1 million people) than expected (the presumption is that whatever these states have done, it is working better than other states). The states highlighted in red are experiencing a higher rate of COVID-19 cases (per 1 million) than expected:

Figure 5: Actual versus Expected COVID-19 Confirmed Case Rates (as of 3 May 2020)

Note in Figure 5 that Arkansas is a non-lockdown state, while Florida implemented its lockdown relatively late in their outbreak (29 days after the first confirmed case). The average lockdown state implemented their policy 21 days after the first confirmed case.

Figure 6 shows a similar table for actual versus predicted COVID-19 fatalities per 1 million people.

Figure 6: Actual versus Expected COVID-19 Fatality Rates (as of 3 May 2020)

Non-lockdown state Arkansas now finds itself on the under-performing (red highlight) list with 25 actual COVID-19 fatalities per 1 million people versus an expected 15 fatalities per 1 million people. On the other hand, non-lockdown state Utah is on good side of the ledger with 31 expected COVID-19 fatalities per 1 million people versus an expected fatality rate of 83 per 1 million people.

Final Thoughts

As of now, the quantitative evidence is mixed in support of full, statewide lockdown policies as being effective in mitigating and suppressing the spread of the coronavirus. To the extent lockdowns matter in controlling the spread of the virus, the analysis presented here cannot distinguish what elements of a lockdown policy are most effective. It could be that merely enforcing safe personal behaviors and social distancing — such as face masks, preventing large group activities, and pressuring people to keep at safe personal distances — are having the most impact. ‘Staying-at-Home’ may be unnecessary.

More importantly, the actual mortality rate of COVID-19 (along with the number of new daily cases) is critical information needed to make a wise decision on statewide lockdowns. Some extremely vulnerable populations — the elderly and those with certain preexisting medical conditions — may need to continue to stay at home; but, for the vast majority of Americans, it may not be a wise policy to continue, particularly in light of the economic damage the lockdowns are having across the globe.

South Korea didn’t lockdown. Japan didn’t. Taiwan didn’t. Singapore didn’t. Hong Kong didn’t. Yet, they are all doing better than the U.S. in COVID-19 fatality rates and they didn’t shut down their economies in the process.

The news media and Democrats may continue to dismiss or censor this type of information and cherry-pick evidence that supports full lockdowns, but doing so stunts what should be an open and constructive debate. [Remember those?]

And, frankly, I have provided some tentative evidence to support statewide lockdowns, though I draw no absolute conclusions on the importance of continuing statewide lockdowns.

The empirical question has not been definitively answered on lockdowns and may not be known until this pandemic is over and epidemiologists and public policy researchers can apply more sophisticated modeling to the final numbers. We will eventually find out the true mortality rate for COVID-19 — not unlike what epidemiologists at The University of Puerto Rico at Mayagüez recently did when they concluded the U.S. news media-promoted claims that up to 4,000 Puerto Ricans died in 2017’s Hurricane Maria were substantially inflated. They determined between between 1,069 and 1,568 Puerto Ricans died directly and indirectly from Hurricane Maria. Some will rightfully claim that number is still too big to be acceptable, but it is substantively different from the misinformation promulgated by the U.S. national media not long after Hurricane Maria hit Puerto Rico’s shores.

Similarly, if we want to avoid learning the truth years from now, instead of now, any suggestion that the debate on lockdowns is settled must be treated as nonsense — and potentially harmful to the country.

• K.R.K.

As always, comments can be sent to: kroeger98@yahoo.com

To my readers that have sent constructive comments and not received a reply from me, I apologize. The volume has increased to the point where I am not able to answer all of them. But I do try.

APPENDIX: The Path Model Estimation Output

Model Set-up in JASP

Model Estimation Output

By Kent R. Kroeger (Source: NuQum.com, May 2, 2020)

My disaffected Bernie Sanders-supporting friends (and other progressive Democrats abandoned by their party) are telling me Joe Biden is unfit for the Democratic presidential nomination due to an alleged sexual assault 27 years ago.

In the #MeToo-era, how can this accusation not be disqualifying for a Democratic candidate?

A fair question.

For me, however, the depth of my opposition to Biden is unaffected by a sexual assault charge that only now, as he stands ready to gain the party’s nomination, comes to light. No person can adequately defend themselves on a accusation that deep in the foggy past. I had the same opinion with respect to Brett Kavanaugh (and lost political friends because of that stance; many of whom now refuse to believe Biden-accuser Tara Reade despite considerably more corroborating evidence for her story than the allegations made against Kavanaugh).

No, my opposition to Biden is rooted in my direct experience with the man 33 years ago (and for which he is equally unable to defend himself).

That is the definition of a hypocrite and I plead guilty on that charge.

Nonetheless, here is my problem with Joe Biden…

I experienced Biden in the 1980s when he had abundant charisma and hair. In his best moments in the U.S. Senate, he was an inspiring orator — equal to Ted Kennedy — who willingly planted himself in the political center and had measurable influence on major pieces of congressional legislation, as many of his more senior, left-leaning colleagues were voted out of office during the Reagan Revolution.

Joe Biden is not blowing smoke when he says, “I’ve gotten things done.”

He really has. Not always good “things,” mind you, but he has a substantial record of legislative accomplishments. That is not my opinion. Look at his legislative record here. Contrast Biden’s lawmaking achievements to any other candidate in the 2020 Democratic nomination race and it is not even close (Minnesota Senator Amy Klobuchar comes in a distant second to Biden).

However, authoring substantive legislation doesn’t win presidential elections. If it were, Ted Kennedy’s 1980 and 1984 presidential victories would have been followed by Jack Kemp’s two-term presidency.

That didn’t happen.

In my research, I have found swing voters most often judge presidential on two factors: Likability and Trust. Competency plays a role in this dynamic, but mostly through its connection to trust.

In this world, Joe Biden should be the perfect candidate.

Until you actually encounter him, as I did during the 1988 presidential race.

_____________________________________________________________

In Fall 1987, I was starting graduate school at The University of Iowa. In a public policy class, taught by the late Professor Russell Ross, we were graced during one session with an actual presidential candidate: U.S. Senator Joe Biden of Delaware.

Biden in 1987 was imposing, with an almost football player-like physique, whose informal rhetorical tendencies were also laced with good substance and meaty ideas. Among the 1988 Democratic Party presidential candidates, only Jesse Jackson was more impressive, and the two front-runners heading into the Iowa Caucuses, Rep. Richard Gephardt (Missouri) and former Massachusetts Governor Michael Dukakis, were proudly on the opposite end of the charisma scale.

In the first Biden campaign rally I attended, he didn’t disappoint. He spoke simply but eloquently on his disapproval of President Ronald Reagan’s Central American policies (Nicaragua, El Salvador) and transitioned effortlessly into a critique of Reagan’s confrontational approach to the Soviet Union (which he ultimately supported).

What I remember most, however, is that Biden was a natural public speaker: a loud, tenor voice (but not too loud) with expressive, disciplined hands. His signature hand move was to form a fist with his right hand and land an imaginary body blow to whatever point he was trying to emphasize. Biden might be Catholic, but in 1987 he spoke with a evangelical minister’s physicality and magnetism.

After over thirty years of studying the speaking styles and body language of political candidates, I still believe the Joe Biden of 1987 was one of the best public speakers I’ve ever seen. He wasn’t Jesse Jackson, but he was better than Ronald Reagan (and I’ll happily go to my grave defending that comparison).

In many ways, Biden was Bill Clinton before Bill Clinton: A centrist Democrat with an open hostility to New Deal–Great Society interventionism and a cagey acceptance of the liberal Democratic social agenda.

If you could have created the perfect centrist Democrat to beat the George H. W. Bush in 1988, it would have been Joe Biden.

Or, so I thought.

In the 30 minutes Biden commanded the attention of my public policy class, something weird happened towards the end — a personality trait that has become conspicuous in the 2020 presidential campaign.

Joe Biden has a hostility problem.

After he gave my class his standard stump speech focusing — as best as I can recall — on the damage the Reagan economy had done to “hard-working families” (as if to contrast with those lazy families), Biden took questions.

Q & As: The wheelhouse for any great politician (Barack Obama, Bill Clinton, John Kennedy).

One of the last questions came from a female classmate who asked a straightforward question about Biden’s support for many of Ronald Reagan’s conservative legislative triumphs, such as the 1981 tax reform bill — that lowered income tax rates, but, according to the non-partisan Congressional Research Service, did little for economic growth and productivity while disproportionately benefiting the wealthiest Americans — and the 1981 Reagan budget bill — that froze, cut and, in some cases, eliminated federal programs for health, education, and other social services. It was a budget that the Washington Post described as “the reversal of two great waves of government intervention, the New Deal and the Great Society.”

The question by my classmate was not overly hostile. “Do you consider yourself a Reagan Democrat?” was essentially her point.

It was the type of question Bill Clinton would have knocked out of the ballpark during his 1992 presidential campaign; and the 2008 version of Hillary Clinton, as prickly as they come, knew how to forcefully address unfriendly inquisitors without finger wagging or insulting their intelligence.

Not Joe Biden.

He bushwhacked his female inquisitor. The specifics of his verbal attack are now fuzzy, but I recall words and phrases like ‘naive,’ ‘lack of experience,’ and ‘live in the real world’ being directed at her.

The problem wasn’t him, it was her!

This answer shouldn’t have surprised me given that one the campaign themes tested by the 1988 Biden campaign was “Scold the Voter.” [I wish I made that up.] Political researcher Paul Taylor, in fact, documents the fratricidal warfare within the Biden camp that allowed such an titanically stupid idea to have life, if only for a brief moment.

Most shocking to me about Biden’s response to this young woman, however, was the personal nature of his answer. Joe went from zero to light speed on the ferocity scale…for no obvious reason! It was a softball question. I can still see her face, her mouth agape, staring at Professor Ross as though about to ask if she said something wrong.

To Biden’s credit, he admitted he had “been too harsh” in his response, which significantly defused the tension. But the damage was done for me.

A well-balanced politician should be able to explain their votes on major pieces of legislation, particularly complicated budget bills that are typically packed with goodies and pet agendas that politicians can sell to their constituents back home.

It was not a hard question — and certainly not personal in tone.

I left that class thinking Joe Biden was odd, in the scary sense. As if he could, in a moment of passion, kill someone and bury the body along the Rock Creek Parkway. [I’m not saying Joe Biden has done that, only that he seems capable.]

In observing Joe Biden’s career-spanning instances of public fury, often manifesting at the strangest times, it is impossible not to think of his personal tragedies as possible sources of this latent anger. I have genuine empathy for Joe Biden, even as I cringe watching him in situations like this one below where he calls a 21-year-old woman a “lying dog-faced pony soldier”:

Admittedly, Biden was trying to be funny in supposedly borrowing the “lying dog-faced pony soldier” quote from a John Wayne movie. [More likely, it was a misquoted line from the movie, “Pony Soldier,” starring Tyrone Power.]

But to me, Biden still comes across as needlessly hostile, just as he did 33 years ago.

I realize I’m judging Biden on something far less serious than a sexual assault allegation; but, for that very reason, I feel comfortable using his unambiguous personality traits as judgment factors. The fact that Biden keeps reinforcing my doubts about his temperament through what he says and does in the present only further validates my vote decision process.

Politics is not a forgiving, merciful business. Politicians can once have had problems, but they can’t have problems. Americans love survivors, not victims.

And, since that day in 1987 when Joe Biden came to speak to my class, I can’t get over the feeling that Joe Biden has significant personality problems.

• K.R.K.

Send comments to: kroeger98@yahoo.com

By Kent R. Kroeger (Source: NuQum.com, April 15, 2020)

Key Takeaways: At this point in the 2009–2010 coronavirus pandemic (9 April), the country-level variable that most strongly correlates with the incidence rate of coronavirus-related deaths is the number of hospital beds relative to population size.

Curiously, or perhaps not, there is also tentative evidence that countries with the most political freedom are experiencing higher numbers of coronavirus-related deaths per 1 million people, all else equal.

This could be a function of such countries having a harder time implementing restrictive, yet effective, mitigation and suppression policies — such as travel bans, stay-at-home orders, and business closures — due to public resistance, or…

…it could be because countries with political freedom are aided by law, economic means and/or cultural norms to be more forthcoming with accurate mortality figures.

And, of course, both statements could be true.

Accurate record keeping on deaths will be critical to our understanding of the coronavirus pandemic of 2009–2010. It is, somewhat sadly, a record keeping function governments do fairly well.

Governing authorities have collected vital records on births, marriages, and deaths for centuries. Among the earliest was China’s Xia Dynasty (2070 BC — 1600 BC), which instituted household registration as part of a larger effort to pacify the Great Floods along the Yellow River that annually devastated Chinese agriculture.

China’s Han Dynasty has the oldest surviving census data in the world, collected in 2 AD when China’s population numbered 57.7 million in 12.4 million households.

When European monarchs and aristocrats during the Middle Ages found such data collection threatening to their privileged status, Church parishes picked up the slack by keeping records on births, marriages and deaths — and in the case of deaths, it was often no more than a ledger entry acknowledging that the deceased’s family paid the Church for burial fees.

More relevantly, the World Health Organization (WHO) has been periodically measuring the completeness of cause-of-death documentation in 112 countries since 1990, with scores ranging from 0 (no cause-of-death information) to 100 (complete information on cause-of-death).

In the most recent WHO data, a little over half of all countries (including most of Europe) achieved a perfect score, and around 70 percent of countries earned a score of 90 percent or better. The countries with the lowest scores (all scoring under 60 percent) were Kuwait, United Arab Emirates, Dominican Republic, Peru, Jordan, Albania, Malaysia, Qatar, and Saudi Arabia.

While there are good reasons to suspect the numbers for national-level confirmed coronavirus cases are not unbiased measures of the virus’ actual incidence within these populations — as the number is also a function of test availability — there should be slightly more confidence that the number of coronavirus-related deaths is a more reliable outcome measure when comparing countries.

As many government experts and social scientists have already warned, the coronavirus numbers being reported by the WHO (for cases and deaths) are not unbiased and are especially difficult to compare across countries. In fact, impossible to compare across countries when there is wide variation in the implementation of mass testing, says economist Derek Scissors.

Nonetheless, there are questions that need to be answered — such as the role of political freedoms in implementing effective coronavirus mitigation and suppression policies — and we have abundant, though flawed, cross-national data to help answer them.

Statistics professors love to warn their students about the dangers of “garbage-in-garbage-out” statistical models. But I consider the dangers of “nothing-in-nothing-out” statistical models to be equally problematic.

It is possible to work with and learn from messy data, if presented with the proper caveats. And such is my attempt here.

The Research Question

What national-level factors independently correlate with the relative number of coronavirus deaths so far experienced by countries worldwide? Factors tested in this analysis initially included:

• National wealth (2019 GDP per capita)
• Income inequality (GINI coefficient)
• Political rights (Freedom House Political Rights Index — PRI)
• Medical capacity (Hospital beds per 1000 people, Doctors per 1000 people, Nurses per 1000 people, Pulmonary Admissions per 1000 people)
• National Demographic and Health Characteristics (Percent Smokers, Percent Overweight, Percent Diabetic, Percent Over 65 years old)
• Geographic location (Latitude, Longitude)
• Presence of a national universal health care system (1 = Yes, 0 = No)
• Coronavirus Testing Frequency (as of 9 April, according to OurWorldInData.org)
• Day Since First Coronavirus-related Death (based on WHO data)

The Dependent Variable

The number of coronavirus-related deaths per 1 million people (as reported to the WHO) as of 9 April 2020. The total population numbers were obtained through The World Bank Open Data portal. Due to this variable’s skewed, non-normal frequency distribution, its logged value was used in the statistical models.

The Countries Analyzed

As so often is the case, it is hard to find complete data for all countries — and even if we did, is it actually helpful to include the Faroe Islands or Nauru when the central inquiry is more about China’s Cultural Revolution-scale virus suppression and mitigation strategies (e.g., hundreds of thousands of volunteers checking people’s temperature, logging their movements, and overseeing quarantines) than if or why small North Atlantic or Pacific islands are able to avoid the coronavirus pandemic?

Therefore, for my analysis here, I limited the countries to those with at least 3 million people which were also categorized as upper-middle-income or upper-income by The World Bank. The three million population cut-off eliminated many of the countries with large amounts of incomplete data.

Within these 91 countries there is significant variation (as of 9 April) in the number of COVID-19-related deaths per 1 million people.

The 10 countries with the most COVID-19-related deaths per 1 million people (as of 9 April) are:

The 210 countries with the fewest COVID-19-related deaths per 1 million people (as of 9 April) are:

After dropping any remaining countries with too much incomplete data on the key variables, I was left with 91 countries for the linear model estimates. I also repeated the model estimates using only the upper-income countries (n = 48), losing most of the non-democratic countries from my original sample (including China and Iran).

Since the parameter and model fit estimates were only marginally different between the two frames (except for one variable — political rights — which will be discussed in more detail), I will only present the linear model results for the 91 countries. Linear model estimates for the 48-country frame are in the Appendix at the end of this article and diagnostics are available upon request to: kroeger98@yahoo.com.

Estimation Methods and Software

All results presented here were conducted using the software package, JASP, a free and open-source graphical program for statistical analysis supported by the University of Amsterdam. Its easy-to-use graphical interface resembles commercial software like SPSS and includes options for standard (frequentist) and Bayesian linear regression models (see Figure A.3 in the Appendix for some of the Bayesian model output).

The Results

Data on COVID-19 is the foundation on which we understand this pandemic and how it is spreading.

Data equips us to respond effectively to this threat, both as individuals and as a society. More importantly, it allows us to learn the best individual-level and societal-level countermeasures to fight against this threat.

The most important data definitions, therefore, are what defines being a carrier of the coronavirus and its role in a final disease outcome (i.e., surviving versus death).

The World Health Organization defines a confirmed COVID-19 case as “a person with laboratory confirmation of COVID-19 infection.” By definition, therefore, a death resulting from COVID-19 requires a positive test result on the coronavirus. This is the assumption I use when modeling coronavirus-related deaths worldwide.

The first step in understanding the relationship between variables is to look at their bivariate correlation. In the case of coronavirus-related deaths we see some interesting correlations in that regard. Figure 1 through 3 show the bivariate relationship between COVID-19 deaths and the number of hospital beds (per 1,000 people), the number of doctors (per 1,000 people), and the Political Rights Index (as measured by the Freedom House).

Figure 1. Bivariate relationship between COVID-19 deaths and hospital beds per 1000 people.

In the case of hospital beds (per 1,000 people), the relationship has some nuance. Overall, there is no clear relationship. But if we focus a subset of countries based upon economic development, there is an apparent connection. Wealthy, high-developed economies with a relatively low number of hospital beds per 1,000 people (e.g. Spain, Italy, UK, Sweden, Switzerland, Belgium, US, Luxembourg, Ireland and France) seem to have a high incidence of COVID-19-related deaths. Whereas, among highly-developed countries with a relatively high number of hospital beds per 1,000 people (e.g, Japan, South Korea, and Russia), the incidence rate of COVID-19-related deaths is much lower.

Figure 2. Bivariate relationship between COVID-19 deaths and Freedom House’s Political Rights Index.

The relationship between political rights and COVID-19-related deaths is similarly nuanced. Within the highly-developed economies (mostly occupying the far-right quadrant of the graph in Figure 2) there seems to be no relationship; however, when adding the upper-middle-income countries, the relationship becomes decidedly positive: countries with the most political freedom are suffering the highest incidence rates of COVID-19-related deaths.

Is this a function of their greater ability of less-free countries to constrain citizen rights in the pursuit of effective coronavirus containment strategies? Or is it a function of the higher quality (and honesty) of data emanating from freer countries?

Figure 3. Bivariate relationship between COVID-19 deaths and doctors per 1000 people.

Lastly, in Figure 3, we have the relationship between the number of doctors per 1,000 people and COVID-19-related deaths. As the relative numbers of doctors in a country is correlated with the relative number of hospital beds, the bivariate plot with COVID-19-related deaths is similarly positive in form. However, we do not see a similar subset relationship within highly-developed economies. Unlike the hospital beds graph, we do not see a cluster of highly-developed economies in the lower-right-hand quadrant of the doctors-per-1000-people graph. At least in the bivariate context, it does appear the relative number of doctors in a country is positively related to the relative number of COVID-19-related deaths.

Can a country with too many doctors per capita cause COVID-19-related deaths? Of course not. That is not the conclusion anyone should draw from Figure 3.

What is true is that upper-income countries are experiencing significantly more COVID-19 deaths per capita than upper-middle-income countries. Thus, predictor variables strongly correlated with national wealth (such as Doctors per 1,000 people and Freedom House’s Political Rights Index) tend to be positively related to the number of COVID-19-related deaths. Those are not necessarily causal relationships.

My hunch as to why the relative number of doctors is positively correlated with coronavirus deaths is that wealthy countries with a high percentage of citizens over 60 years old will tend to have more doctors (per capita). Those most vulnerable to the coronavirus (the elderly) are also more likely to live in an advanced economy.

That is what I think is going on.

We learn more when we consider all of the predictors jointly in a linear regression context.

Figure 4 shows the linear model estimates for COVID-19-related deaths using the following independent factors for each country: (1) its latitude (at its geographic center), (2) its political rights index score, (3) its incidence of doctors (per 1,000 people), (4) its incidence of hospital beds (per 1,000 people), (5) its GDP per capita, and (6) the number of days since its first COVID-19-related death (log transformed).

Along with GDP per capita, other independent factors that were not significant predictors of COVID-19 morality rates included: the presence of a universal health care system and the relative frequency of coronavirus testing. Additional, more dynamic, models will need to be tested to determine the how those otherwise important variables might impact COVID-19 mortality rates.

Figure 4. Linear model for COVID-19-related deaths per 1 million people (log transformed) for upper- and upper-middle income countries with at least 3 million people (n = 91).

Using the standardized beta coefficients in Figure 4 as an indication of the relative importance of each independent factor, it appears the number of hospital beds (per 1,000 people) is the most importance correlate with COVID-19-related deaths (standardized beta = -0.521), followed by days since first COVID-19-related death (standardized beta = 0.512), political rights (standardized beta = 0.490), latitude (standardized beta = 0.445), and doctors per 1,000 people (standardized beta = 0.445).

The overall linear model offers a decent fit to the data with 72 percent of the variance explained.

[Casewise diagnostics for selected countries are available in Figure A.2 in the Appendix below]

Figure 5 illustrates the impact of standard deviation changes in the predictor variables in which the countries have some control over the factor (i.e., a country can’t changed its geographic location).

Figure 5. How standard deviation changes in the “controllable” predictor variables relate to the number of COVID-19-related deaths.

The most notable finding in Figure 5 is how a relatively modest increase in hospital beds per 1,000 people can have a meaningful impact on the number of COVID-19-related deaths (per 1 million people). An increase of 2.38 hospital beds per 1,000 people would reduce a country’s relative COVID-19 death totals by 2.5 people; a 4.4 hospital bed (per 1,000 people) increase would translate into 6.5 fewer COVID-19-related deaths (per 1 million people).

Keep in mind, this is a static cross-sectional (one-point-in-time) model where each country enters the model at a different stage in the coronavirus process. China and South Korea passed through the peak stages of the crisis over a month ago, while Italy and Spain reached their peaks a few weeks ago, and the U.S. is only now moving past the peak of its crisis. For this reason, the variable — days since the first COVID-19-related death — is employed to control for these stage differences. It is an imperfect control, to be sure.

Ideally, the estimated model would be constructed in a dynamic, time-series context in which predictor variable interactions could also be tested for significance. Quite likely, some regression parameters will vary by country-types, for example.

However, for the sake of simplicity, the static model reported here still reveals the relative importance of the independent factors, even without the full dynamic model.

Figures 6 and 7 compare linear model estimates to the actual values for COVID-19-related deaths. Figure 7, specifically, filters down to the major countries in this analysis.

Figure 6. Actual versus Predicated Values from a Linear Model for COVID-19-related Deaths per 1 million people.

Figure 7. Actual versus Predicated Values from a Linear Model for COVID-19-related Deaths per 1 million people (Selected Countries Only).

One rule-of-thumb in outlier identification is to look for cases with a Cook’s D of greater than 3. As seen in Figure A.2 in the Appendix, China has a Cook’s D value of 38. China is an obvious outlier and a good reason for re-running the linear model using only upper-income countries (which excludes countries like China and Iran and leaves us with only 48 cases).

However, by excluding countries like China and Iran, the statistical significance of Freedom House’s Political Rights Index disappears (probability of t-value = 0.154, two-tailed test), for the likely reason that when eliminating upper-middle-income countries from the analysis we lose much of the variation in the Political Rights Index.

Still, if we split the 48 countries into two groups — Countries with High Political Rights Index scores (PRI > 33) and Countries with Lower Political Rights Index scores (PRI =< 33), a group that includes the U.S., Israel, Hungary, South Korea, Bahrain, Brunei, Qatar, UAE, Saudi Arabia, Singapore, Oman and Kuwait — we can compare the mean number of COVID-19-related deaths (per 1 million people) for the two groups.

The low political freedom countries had an average number of COVID-19-related deaths of 6.9 versus 50.8 for the high political freedom countries. And, as it turns out, the t-statistic for the difference-in-means test (t = 1.718, p = 0.093, df = 46) is still not significant.

Nonetheless, I believe the insignificance of the political rights factor is one of methodology and sample size and not its inherent importance. With a more systematic approach, I expect we will see the significant role a country’s political rights had on how they responded to the coronavirus pandemic.

Final Thoughts

Clearly missing in the linear model results presented here are controls for the types and timing of coronavirus mitigation and suppression policies implemented by each country. Some countries adopted broad, systematic testing protocols early (South Korea, Taiwan, Singapore, Germany) and the end benefit has been that they’ve experienced fewer deaths per capita than countries that came to systematic, mass testing late.

CNN’s recent story on the success of German efforts to control the coronavirus reinforces the findings from the model reported here. Writes CNN’s Angela Dewan, Henrik Pettersson and Natalie Croker:

The difference appears to be the German healthcare system’s huge capacity. Germany is projected to need about 12,000 beds at the peak of this outbreak in the middle of the month, according to projections from the Institute for Health Metrics and Evaluation. It has over 147,000, more than 10 times its need.

In contrast, the US has around 94,000 beds, some 15,000 beds short of its need. Germany has more spare beds in intensive care units that Italy has altogether.

In fact, Germany’s health system has such a large capacity, its hospitals are now treating people for coronavirus from Italy, Spain and France.

Hospital beds are merely a proxy for the real factors that have saved lives during the coronavirus pandemic of 2019–20. With a hospital bed comes the hour-by-hour care of nurses and other health professionals. There is no substitute for a hospital bed. The fact that more New York City residents than usual are dying at home due to the coronavirus should not be a surprise to anyone.

It is all about capacity. Do you — city, state, or country — have enough hospital beds to handle a surge in need? Figure 1 (above) reveals this simple reality. South Korea and Japan (and to a lesser extent, Germany and Austria), have enough hospital beds to handle a crisis like the coronavirus.

You can’t ramp hospital beds late in the game (as China did) and expect to escape a virus’ rampage. There must be planning. As World War II Allied Commander, General Dwight Eisenhower, once said: “In preparing for battle I have always found that plans are useless, but planning is indispensable.”

Such is the lesson of the 2019–20 coronavirus pandemic.

But there are other lessons.

My Russian-born friend, Zhora, an American citizen, recently told me, “Americans do not know how to be sick.”

Americans work when they are sick. They cough on you without consideration or remorse. They are working, after all. Why should they be sorry? Americans wear their illness among workmates like a badge. People at home sick are not working. What good are they? They aren’t getting the big promotion.

A new economic world is forming under our feet. Workaholics aren’t the modern gladiators they pretend to be. Since the coronavirus pandemic, they are genuine threats to the public good.

This is a preliminary analysis. For one, the coronavirus (2019-nCoV) and its associated disease (COVID-19) are still working through the world’s population. And secondly, explaining COVID-19-related deaths requires a dynamic modeling approach (i.e., references to time) to fully understand the process by which this pandemic has played out worldwide so far.

Stay tuned…and stay home, for now.

• K.R.K.

Datasets used in this analysis can be obtained by request to: kroeger98@yahoo.com

APPENDIX

Figure A.1. Linear model for COVID-19-related deaths per 1 million people (log transformed) using only upper-income countries (n = 48).

Figure A.2. Casewise diagnostics for linear model of COVID-19-related deaths per 1 million people (log transformed) for upper- and upper-middle income countries with at least 3 million people (n = 91).

Figure A.3. Posterior distributions and credibility intervals for Bayesian regression coefficients in a linear model of COVID-19-related deaths per 1 million people (log transformed) for upper- and upper-middle income countries with at least 3 million people (n = 91).

By Kent R. Kroeger (Source: NuQum.com, April 12, 2020)

“The government are very keen on amassing statistics. They collect them, add them, raise them to the nth power, take the cube root and prepare wonderful diagrams. But you must never forget that every one of these figures comes in the first instance from the village watchman, who just puts down what he damn pleases.”

— Josiah Stamp (1880–1941), an English industrialist, economist, civil servant, statistician, writer, and banker

Barely a month into the coronavirus pandemic, the congressional Democrats and Republicans are already viewing this human tragedy through their narrow, partisan lenses.

And it was only two weeks ago they were working so well together, when the Congress passed and President Trump signed the $2 trillion bailout bill — U.S. history’s greatest wealth transfer from the U.S. Treasury to wealthy Americans.

That is what Democrats and Republicans can agree on when the American people are preoccupied with whether to go to the grocery store to re-stock their water and food supply, thereby risking coming down with the scariest virus since the fictional Andromeda Strain.

Now that both political parties have guaranteed that wealth inequality is baked into the American economic system for at least another generation, they can get back to pretending they can’t agree on anything.

And what is the current partisan disagreement de jour? Whether the Americans should go back to business-as-usual, or continue to shelter-in-place until every last known RNA segment of the coronavirus has been wiped from the face of the earth — or, at least, the blue states.

While the bailout bill guarantees the Top 1% will survive the pandemic with their bank accounts intact (plus a hefty Christmas bonus), a significant percentage of Americans are genuinely at risk of going financially bankrupt if this pandemic lockdown extends beyond summer. Even Trump understands those people represent voters who could potentially tip the balance against him in November. The Democrats, likewise, led by Alexandria Ocasio-Cortez and the Cuomo brothers, Governor I’d-Better Cover-My-Backside and Fredo, see a tremendous opportunity to extend the Blue Wave started in the 2018 midterm elections.

The Republicans, for good reason, want the U.S. economy back to normal as quickly as possible. Every day this near-national lockdown extends, the more Republicans, independents, and disaffected Democrats who are adversely affected and potentially more likely to vote against Trump.

On the other side, the Democrats are saying if end this lockdown too soon we risk a second wave of the coronavirus. And what is too soon, according to the Democrats? Anything that ends this economic catastrophe before November is my guess.

Most disturbing about this morbid ballet dance now going on in Washington, D.C. and in the national media is how both sides cherry pick the coronavirus numbers that best support their policy position. There is nothing new about that. Politicians do that on every issue.

But it is particularly damaging in the case of coronavirus pandemic because it indeed puts lives and livelihoods at additional risk. Furthermore, it perpetuates the myth that the statistical numbers governments and independent agencies are producing for decision making about the coronavirus are accurate…when, in fact, they are not.

That is not an assertion on my part. It is a well-documented characteristic of the statistics gathered during previous epidemics and pandemics. The numbers initially collected and tabulated are invariably under-counts of the true spread and mortality rate of a virus.

The 2009 swine flu pandemic is a perfect example. In August 2010, over a year after the pandemic started, the World Health Organization (WHO) had confirmed 1,632,710 cases of the swine flu worldwide and 18,449 resulting deaths. The WHO knew it was an under-count and reported the numbers explicitly with that caveat.

It would not be until August 2011 and June 2012 that the WHO and U.S. Centers for Disease Control (CDC) would release their modeled estimates of the 2009 swine flu’s actual reach worldwide: 700 million to 1.4 billion people infected and between 151,700 and 575,400 dead.

And those modeled estimates too are subject to error and dispute.

So why is the news media telling us the U.S. has had six times more confirmed cases of COVID-19 than China (530,200 to 83,134 at last count), when they know (and we know, and even China knows!) that isn’t close to true.

Some estimates put China’s COVID-19 cases and deaths at up to 10 times the government-sourced numbers, including my own estimate. And it is not just China’s numbers being questioned. The Italian government’s own civil protection authority has estimated Italy’s true case numbers are 10 times higher than officially reported, and for many of the same reasons used to dispute China’s numbers, including: (1) lack of testing kits, (2) unmeasured asymptomatic cases, (3) symptomatic cases who never sought medical help, (4) misclassification by health officials (e.g., failing to properly categorize COVID-19-related pneumonia cases, (5) government changing definitions for COVID-19 classification, and (6) simple human error.

Yet, there is no reason to pick on China and Italy. The U.S. numbers most likely are equally wrong — and for the same reasons.

Of course, Fox News and the conservative blogosphere have found reasons to believe the U.S. is over-counting coronavirus cases and deaths, noting an apparent drop this year in CDC reported pneumonia-related deaths, compared to previous years.

“For the last few weeks, that [pneumonia] number has come in far lower than at the same moment in previous years. How could that be?” Fox News host Tucker Carlson asked on his show recently.

The Washington Post’s Aaron Black rightly noted that the CDC pneumonia-related death numbers were not down by much and historically those CDC-reported numbers tend to be under-counts (that word again!) when the numbers are first reported. The CDC will produce more accurate numbers later in the year (and take their word for it, those numbers will be right!).

While Trump’s own director of the National Institute of Allergy and Infectious Diseases, Dr. Anthony Fauci, also knocked down that Fox News theory, it is a reasonable assumption that some pneumonia-related deaths have been classified as coronavirus-related despite no lab-test confirmation (recall Reason #1 above for why national coronavirus counts are likely wrong).

The question shouldn’t be if this has happened, but, how frequently this has happened. And asking that question is not tacit support for Fox News’ conspiracy theory, it is merely an acknowledgement that these coronavirus statistics contain error.

In my introductory statistics classes I always shared with students what data scientists call The Confusion Matrix:

It shows the potential for error in any measurement process. In this example, the matrix is updated to apply to the COVID-19 pandemic.

If COVID-19 existed in a perfect world, every government would accurately measure every citizen for the coronavirus on a regular basis. Everyone would either fall into the True Positive category or the True Negative category.

That is not happening…not in the U.S….not in China…not in Peru…not in Germany…not even in South Korea (the apparent gold standard on how to measure and contain the coronavirus).

The vast majority of world citizens have not been tested for the coronavirus and probably will never be tested. According to OurWorldInData.org, 8.3 million tests (as of 8 April) have been reported by government authorities worldwide. But we already know that number is low because some countries — particularly outside Europe, North America and East Asia — have not reported their testing numbers (including China).

But even if we double the testing number, only about 0.2 percent of world citizens have been tested for the coronavirus — and that number is squishy because countries don’t always distinguish between the number of tests they’ve administered (some people require more than one test) versus the number of people tested.

Keeping in mind The Confusion Matrix, Chinese scientists report preliminary evidence that the most common type of COVID-19 test — the reverse transcriptase polymerase chain reaction (RT-PCR) test — may give false-negative results around 30 percent of the time.

Dr. Harlan M. Krumholz, a professor of medicine at Yale University and director of the Yale New Haven Hospital Center for Outcomes Research and Evaluation thinks the U.S. false-negative rate is even higher.

”Unfortunately, we have very little public data on the false-negative rate for these tests in clinical practice,” says Dr. Krumholz. And there are false-positive tests cases as well — though our better-safe-than-sorry bias makes us less concerned about those cases.

In sum, of the six disposition cells in The Confusion Matrix, four work against the accuracy of any number being generated about the coronavirus. We can minimize how many people fall into these cells, but can’t empty these cells entirely.

This is why modeled estimates are so critical to understanding the coronavirus and its associated disease, COVID-19. They explicitly acknowledge classification error and try to compensate for it. The models aren’t perfect, God knows can I attest to that, but they are typically better than the raw data when it is initially collected.

Which is why the political debate right now about when the U.S. can return to normal (which I don’t believe will happen until there is a vaccine) is so unproductive; and, indeed, does harm to the general public’s understanding of the coronavirus.

The Trump administration, congressional Republicans, and their loyal media outlets, are starting to minimize the severity of the coronavirus and risk re-igniting its spread if they follow through on plans to return the U.S. to near business-as-usual too soon (in May if Trump gets his way).

Likewise, congressional Democrats have cherry-picked worst-case scenario forecasts (which are fast being proven wrong, particularly predictions about the U.S. death toll) that do more to create panic in the general public than build their knowledge about the coronavirus.

Underwriting the fact-denial and fear-mongering are the coronavirus numbers which offer evidence for both sides. But the reality is that we still don’t know how far the coronavirus has spread within the population (morbidity rate), its contagiousness (R0 — “R naught”), or its mortality rate [with estimates ranging from the “common flu” (0.1%) to SARS-like numbers (10%) in some locations].

As long as we don’t have a better handle on these numbers, our political debates are going to be disproportionately driven by the hourly updates on Johns Hopkins University’s COVID-19 Dashboard website. That is a bad thing because these numbers are not accurate.

[According the COVID-19 Dashboard, the number of U.S. coronavirus cases just rose from 530,000 to 542,000 while I am writing this article.]

Economic research blogger, Christopher Balding, puts it best: “The panic and debate clearly suffers from profound misunderstandings.”

That is probably another under-estimate.

• K.R.K.

Datasets used for this article are available upon request to: kroeger98@yahoo.com

By Kent R. Kroeger (Source: NuQum.com, April 9, 2020)

Key Takeaways: There is no systematic evidence that the overall quality of a country’s health care system has had an impact on the spread (morbidity rate) and lethality (mortality rate) of the coronavirus. Instead, a country’s per capita wealth and exposure to the international economy (particularly international tourism) significantly increases the spread of the virus within a country. This latter finding may be partly a function of wealthier populations being more likely to have their coronavirus-related illnesses diagnosed and treated. But it is also likely that international travel is spreading the virus worldwide.

As for the mortality rate, the story is more complicated: The single biggest driver of the mortality rate, so far, is simply the time since the country’s first coronavirus-related death. Once the virus has found a vulnerable host, the final outcome may be difficult to change (at least for now).

As for the charge by the US intelligence community that China has under-reported the coronavirus’ severity in their country, the model reported here suggests China, given its size and characteristics, should have so far experienced 10 times the coronavirus cases they have reported and a case fatality rate twice their current estimate. If they are under-reporting, as charged by the US, China may have between 33,600 to 70,000 deaths related to the coronavirus, not the 3,339 they are currently claiming.

To the contrary, it is also plausible that their aggressive suppression and mitigation efforts have successfully limited the spread and lethality of the coronavirus. The model reported here cannot determine which conclusion about China is true. Or if both conclusions have truth.

_________________________________________________________________

Its OK to feel some tentative optimism about the coronavirus pandemic. It does appear, finally, that the virus and its associated illness — COVID19 — is peaking in many of the countries hardest hit by the virus (see Figure 1).

Figure 1: New daily COVID-19 cases in Italy, South Korea, Iran and Spain

Almost a month-and-a-half after the coronavirus reached its peak in new daily cases in South Korea (around 900 cases-a-day), the virus has peaked in Italy around March 22nd, and in Spain and Iran around April 1st.

If President Donald Trump’s advisers were correct in Monday’s White House daily coronavirus update, the U.S. may also witness its peak in new daily cases within the week.

This weekend, New York, the current locus of the US outbreak, saw a significant decline in the number of new infections and deaths.

“In the days ahead, America will endure the peak of this pandemic,” Trump said Monday.

In fact, from April 6th to 7th, the aggregate US data showed its first day-to-day drop in the number of new COVID-19 cases since late March (see Figure 2).

Figure 2: Cumulative and new daily COVID-19 cases in the U.S.

In many of US states hardest hit by the coronavirus — such as New York, Washington, and California — the number of new cases each day have leveled off or declined in the past week.

These are genuine reasons for optimism. While Trump’s hope for an economic return to near-normal by Easter was overly optimistic, the possibility it could happen in early May is not.

Europe and the U.S. were caught flat-footed by the coronavirus, but it is looking increasingly like they will escape with far fewer cases and deaths than originally anticipated by many epidemiological models.

[Of course, additional waves of this virus may still occur and we may never see a true return to normal until a coronavirus vaccine is made widely available — and by widely available I mean free to everyone.]

________________________________________________________________

In this moment of cautious cheer, my questions increasingly focus on how the world measured (and mismeasured) this pandemic and what national-level factors may have suppressed and, conversely, aided the spread of the coronavirus?

Everyone has theories. Some are convinced autocratic countries (i.e., China, Iran, Venezuela, Russia) have hidden the true impact of the coronavirus on their countries. Others have declared the coronavirus proves the importance of universal health care in containing such viruses. Still others have conjectured the number of COVID19-related deaths have been over-reported by anti-Trump forces, most likely to make Trump look bad. Conversely, the national media has unofficially declared (without conclusive evidence, as usual) that the US government has been under-counting COVID19 deaths (presumably to make the Trump administration look more effective in its coronavirus response than justified).

It is speculation at this point. It will be many months — probably years — before we know what actually happened during the 2019–20 Coronavirus Pandemic. The coronavirus pandemic is still on-going, after all, and the reality is: counting the number of people with any disease or virus is genuinely hard and prone to human error.

But we can start to address some of the controversies, if only tentatively.

If we assume that the majority of countries have exercised a fairly high level of due diligence in measuring the presence of the coronavirus within their jurisdiction, we may be able to identify those countries who have been much less than honest.

Moreover, after controlling for suspected dishonest coronavirus measurement, we may also see hints at the impact of national health care systems and containment policies on the spread and lethality of the coronavirus.

________________________________________________________________

Let us start our inquiry with this premise — there are two fundamental measures of the coronavirus: (1) the number of confirmed coronavirus cases relative to the total population (morbidity rate), and (2) the number of coronavirus-related deaths as a percent of those confirmed to have the virus (mortality rate).

For simplicity’s sake, what I am calling the mortality rate is actually the case fatality rate. In reality, the coronavirus’ mortality rate is much lower than the case fatality rate as its calculation will include undiagnosed cases experiencing only minor or no symptoms.

If universal health care were ever to show its value, now is the time. The logic is simple: Countries where citizens do not need to worry about the cost of a doctor visit, the probability these citizens get tested and treated early for the coronavirus is significantly higher.

Also, countries with universal health care may also be more likely to institute broad-based coronavirus testing, thereby identifying asymptomatic super-spreaders of the virus. Subsequently, when diagnosed with the virus, these citizens will be isolated sooner from the healthy population. Furthermore, early diagnoses of the coronavirus may also improve the chances infected individuals survive the virus.

Can we see this in the data?

________________________________________________________________

Figure 3 (below) is produced directly from World Health Organization (WHO) data. The chart shows the morbidity rate of COVID-19 (i.e., frequency of COVID-19 cases per 100K people) compared to its mortality rate (i.e., deaths per confirmed case).

I’ve segmented the chart in Figure 3 into four quadrants, each defined by countries’ morbidity and mortality rates. Countries with high morbidity and mortality rates are in the upper right-hand quadrant of Figure 3 (e.g., Italy, France, Spain, Netherlands, UK and Iran.); while countries with low morbidity and mortality rates are in the lower left-hand quadrant (e.g., Russia, Japan, Pakistan, Nigeria, and India).

Figure 3: COVID-19 Cases per 100K persons versus Number of Deaths per Confirmed Case.

What does Figure 3 tell us? In truth, not much.

Ideally, a country would want to be in the lower left-hand quadrant (Low/Low) of Figure 3, right? But a simple inspection of the quadrant reveals it is occupied mainly by countries in eastern Europe, Africa, South America and southern Asia (Russia, Ukraine, Pakistan, India, Nigeria, among others) — few of which find themselves ranked by the WHO among the countries with the best health care systems. One reason for their favorable performance so far may be that the coronavirus hasn’t significantly spread to those countries yet — after all, many are in the southern hemisphere.

Here are two fair questions to ask: Are these countries performing relatively well with the coronavirus due to favorable circumstances (fewer people traveling to and from coronavirus sources like China; climatic context; stronger containment policies — an area where authoritarian governments may have an advantage; and/or better health care systems)?

Or, are some of these countries simply not deploying the resources and expertise necessary to measure the impact of the coronavirus? Do they even have the capacity to do so?

________________________________________________________________

Figure 3 begs more questions than it answers, but it still may hint at some tentative conclusions. For example, experience tells me countries clustered around the intersection of the average country-level morbidity (34 cases per 100K people) and mortality rates (3.4%) are in the accuracy ballpark. If I am feeling generous, that list includes the US and China, along with countries like South Korea, Poland and Turkey.

The countries that raise my eyebrows are the major outliers from the center cluster: Italy, Spain, UK, France, Bangladesh, Nigeria, Indonesia and India.

The variation in the coronavirus mortality rate ranges from 12 percent in Italy to near zero percent for New Zealand (a country with 1,239 confirmed cases and only one death). What could possibly explain this difference in the coronavirus mortality rate between two advanced economies? Could it be their health care systems? WHO ranks Italy’s health care system 2nd in the world, while New Zealand’s is only 41st. Russia has a reported coronavirus mortality rate of 0.8 percent and has the 130th best health care system in the world, according to the WHO.

More in line with expectations, Germany, a country given significant positive coverage for its coronavirus response — plaudits comparable to perhaps only South Korea’s — has a reported 2.1 percent mortality rate on a base of 113,296 confirmed cases.

Why such discrepancies in reported mortality rates?

Dietrich Rothenbacher, director of the Institute of Epidemiology and Medical Biometry at the University of Ulm in Germany, credits Germany’s broad-based, systematic testing as being the reason his country’s mortality figures are hard to compare to other countries.

“Currently we have a huge bias in the numbers coming from different countries — therefore the data are not directly comparable,” Dr. Rothenbacher recently told the BBC. “What we need to really have valid and comparable numbers would be a defined and systematic way to choose a representative sampling frame.”

This is where statistics — my profession — becomes critical. As Dr. Rothenbacher asserts, Germany would not have understood the extent of the coronavirus crisis without testing both symptomatic and asymptomatic cases, just as South Korea and, sadly, only a few other countries have done.

Systematic random sampling needed to be a component of every nation’s coronavirus testing program.

It wasn’t.

In New Jersey, where I live, the office of the state’s Health Commissioner told me I couldn’t get tested for the coronavirus without meeting one of the following qualifications (…it felt like a job application):

• Already being hospitalized and showing symptoms of COVID-19.
• A health care worker showing symptoms and having who been exposed to others known to have the virus
• Anyone known to be part of a cluster outbreak (one example being a recent Princeton, NJ dinner party where multiple attendees were diagnosed with the coronavirus)
• And vulnerable populations (e.g., nursing home residents).

Someone like me, a 55-year-old male with no underlying health problems but showing mild flu symptoms — low-grade fever, persistent cough, and chest congestion — cannot get tested in New Jersey.

The New Jersey testing protocol is common across the U.S. given the relative scarcity of testing kits.

________________________________________________________________

Anytime the anecdotal evidence is contradictory or unclear, I turn to data modeling — even if crude — to test some of the initial hypotheses surrounding a controversy.

The challenge with the coronavirus is the availability and data quality of the key causal factors we’d like to test in a coronavirus model for morbidity and mortality rates. In the following linear models, I tested these independent variables:

(1) GDP per capita,

(2) Health care system overall quality (according to the 2019 edition of the CEOWORLD magazine Health Care Index),

(3) Quality of health care professionals (also measured by CEOWORLD magazine),

(4) Exposure to foreign tourism (as measured by World Bank’s measure of international tourist arrivals and international tourism receipts as a percent of total exports),

(5) Exposure to East Asian trade (measured as the percent of total merchandise exports),

(6) Percentage of population over 65 years old (as measured by the World Bank),

(7) Hospital beds per 1,000 people (as measured by the World Bank),

(8) Total COVID-19 tests per 1,000 people (as measured by OurWorldInData.org), and

(9) an indicator variable for autocratic countries the U.S. government or U.S. news media have accused of under-reporting COVID-19 cases and deaths (i.e., China, Russia, Iran, Vietnam, Turkey and Venezuela).

Since countries come into the coronavirus pandemic at different points in time and are in different stages at any point in time, I also controlled for (10) the days since the first coronavirus case was confirmed, and (11) the days since the first coronavirus death occurred.

Out of necessity, I limited the data analysis to countries with reliable data on all key independent measures and with populations over 3 million people, leaving the analysis with 76 countries.

[Note: The linear models, however, were not weighted by country population size. For example, China weighted the same as Serbia in the following models.]

The estimated linear models for morbidity and mortality rates are reported in the Appendix below.

Figures 4 and 5 show the model predictions for each country versus the actual morbidity and mortality rates. In the morbidity model graphic (Figure 4), I only show a selection of key countries in order to simplify the data presentation.

Figure 4: Predicted versus Actual COVID-19 Cases per 100K Persons for Selected Countries (as of 4 APR 2020).

Figure 5: Predicted versus Actual COVID-19 Deaths per Confirmed Cases (as of 4 APR 2020).

On the issue of autocratic countries (who are also U.S. adversaries), there is circumstantial evidence that Venezuela, China and Russia have fewer COVID-19 cases than we would expect given their key characteristics, even while their deviance as a group is not statistically significant.

For example, China may have 10 times the coronavirus cases they have officially reported and a mortality rate twice their current estimate. If true, China may have between 33,600 to 70,000 deaths related to the coronavirus, not the 3,339 they are currently claiming.

Likewise, Russia may have 19,500 coronavirus cases, not the 10,031 they have reported to the WHO and Venezuela may have 1,625 cases, not 167 cases.

Even if, according to the model, the reported numbers for China, Venezuela and Russia are low, we can’t rule out the possibility they are low because these countries have done a superior job containing the virus.

Perhaps the most puzzling (and saddest) case is Iran. Our model suggests Iran has experienced far more COVID-19 cases than we would expect given its characteristics. The most recent WHO numbers for Iran are 66,220 confirmed cases and 4,110 deaths.

Has Iran done an especially poor job of containing the virus or are they measuring more comprehensively than other countries? Unfortunately, my model can’t settle that point.

Final thoughts

I anticipated when I started looking at the coronavirus in 76 countries that the quality of their health care system s— starting with affordable, universal health care — would show up as a significant factor in distinguishing between countries that successfully took on the coronavirus pandemic (e.g., South Korea, Germany, Singapore, and Japan) and those less successful (e.g., Italy, Spain, France, UK and Iran).

While the number of hospital beds per 1,000 people does correlate significantly with lower mortality rates (see Appendix, Figure A.2), the overall quality of a country’s health care system did not. In fact, countries with the best compensated medical professionals actually have higher coronavirus mortality rates.

The coronavirus has hit Europe (and China) the hardest. In Italy, the high percentage of elderly helps explain its high volume of cases, but that can’t be the only explanation. And isn’t just that advanced economies have put more effort into measuring the occurrence of the virus in their communities that explains this fact. The coronavirus has found disproportionately more friendly hosts in these societies. We may have to accept that the coronavirus is one of the evolving risks associated with high disposable incomes and deeps global connections through trade and tourism.

I know this: I will never go on a cruise ship ever again.

Theories on why some countries handled the pandemic better than others are also plentiful. The most compelling analysis may have occurred while the pandemic was just starting.

Writing in early March, Chandran Nair, founder and CEO of the Global Institute for Tomorrow, may have come up with the best explanation still. “Strict and centralized enforcement of lockdowns, quarantines, and closures are the most effective way to contain the virus,” wrote Nair. “What’s emerged from the coronavirus crisis is the fact that some states are equipped to handle this type of action, and some are not — and it has little to do with development status.”

Or, more cynically, could we conclude that one of the costs of emphasizing individual freedom is that when collective action is necessary — including a strong, central state response — Europeans and Americans answer the call by hoarding toilet paper and Jim Beam?

I’m not quite there yet. For one, I don’t believe Nair fully appreciates how the modern state and elites consolidate their power during these uncertain times, and how this can leave even more people vulnerable economically and physically to the next pandemic — and there will be another one. Second, for every example of state power getting this done quickly and efficiently, there are dozens more where greed, incompetence, and arrogance lead the state to do more damage than good. Before we give the modern state more power, let us think this through some more first.

Here is what our governments and scientific community should be doing…

If this global pandemic ends relatively soon — as it appears it might — our governments and health researchers must immediately resolve themselves to understand how many people really did get infected by the coronavirus and how many actually died from its consequences.

Currently, we have a global mish-mash of epidemiological data of unknown quality or generalizability. Only probability-based sample studies can give us the real numbers and it is only with those numbers that we can really sit down and decide: What worked and what was a total waste of time and resources?

• K.R.K.

Data used in this article are available by request to: kroeger98@yahoo.com

APPENDIX: The Linear Models

Figure A.1: Linear Model for Confirmed COVID-19 Case per 100K Persons (Morbidity)

Figure A.2: Linear Model for COVID-19 Deaths per Confirmed Cases (Mortality)

By Kent R. Kroeger (Source: NuQum.com, April 2, 2020)

According to recent a Pew Research Poll, 59 percent of Democrats (and those Democrat-leaning) said the coronavirus outbreak is a major national threat, but only one-third of Republicans (and those Republican-leaning) had the same view.

Partisans in the U.S can’t even leave politics out of a worldwide pandemic. Perhaps German writer Thomas Mann was right when he told a German audience in 1930 that “politics is everything.”

Still, we can’t all agree the coronavirus pandemic is a national crisis?

Nope, apparently not.

And our hyper-partisan political culture is a having a tangible impact on coronavirus suppression and mitigation strategies at the state-level.

For example, 37 governors have issued statewide stay-at-home orders, according to CNN. Notable, however, are those states that not issued such a statewide order: Alabama, Arkansas, Iowa, Missouri, Nebraska, North Dakota, Oklahoma, South Carolina, South Dakota, Texas, Utah and Wyoming.

Every one of the stay-at-home laggard states was won by Donald Trump in the 2016 election. Furthermore, among governors in Trump-states that have implemented stay-at-home orders, many have Democratic governors: Kansas, Michigan, North Carolina, Pennsylvania and Wisconsin.

The state that surprises me the most is Iowa — my home state. While the Hawkeye State is not immune to partisan rancor, the state has a long history of leading on important socially progressive policies (public schools, racial integration, Head Start programs, marriage equality, etc.).

Not anymore.

“I can’t lock the state down,” Iowa Governor Kim Reynolds told reporters last Tuesday. “I can’t lock everybody at home.”

But, actually, Iowa law says she can — she just chooses not to.

According to the Iowa governor, the “internal data” she is looking at concludes that a stay-at-home order is not necessary.

What “internal data” is she talking about? She isn’t clear on that. Is it a super secret data modeling method that forecasts coronavirus transmission? Who knows?

And, while I’m picking on Iowa, a big ‘Boo’ on the Iowa press corps for not pressing Reynolds on exactly what “internal data” she is using to make this life-and-death decision. By the time Reynold’s secret data source signals a stay-at-home order is necessary, it will be too late for hundreds of Iowans who will die because her short-sighted response to the crisis.

Says Reynolds about her reluctance to issue a stay-at-home order, such a policy would put undue strain on the supply chain and essential workforce.

Is Iowa’s economy more important than its retiree population? Reynolds might not put it exactly that way, but by delaying the stay-at-home order (and she will issue such an order at some point), that is precisely what she is saying.

With one of the oldest populations in the U.S., Iowa is particularly susceptible to a coronavirus outbreak in nursing home and long-term care facilities. In fact, Reynolds shared some of her coronavirus data when she admitted a Cedar Rapids (Linn County) long-term care facility — Heritage Specialty Care — is already experiencing an “outbreak” with 30 confirmed cases.

As the tragic experience at the Life Care Center in Kirkland, Washington attests, once an outbreak starts within a vulnerable population living in close quarters, the virus is impossible to stop and the results are deadly.

If Reynolds’ internal data isn’t telling her that “Iowa is a ticking time-bomb,” then it is junk data.

__________________________________________________________________

Now that I’ve vented my frustration about Iowa, here is the really distressing thing I’ve discovered in the U.S. data regarding the coronavirus: There is clear evidence that the extent to which states are testing for the coronavirus is significantly impacted by partisan politics.

I get it that some people believe the ‘coronavirus panic’ is more hype than reality. Why didn’t the world lock-down during the 2009 Swine Flu pandemic that killed 150,000–575,000 worldwide, including approximately 12,000 Americans, according to a Centers for Disease Control (CDC) estimate?

It is a legitimate question, though Dr. Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, aptly responds with three reasons the coronavirus is different from the 2009 Swine Flu: (1) The 2009 swine flu pandemic was spread out over many months and didn’t place nearly the level of strain on the world’s health care system that the coronavirus has done in just a matter of a weeks; (2) the medical community has extensive experience developing vaccines for influenza viruses and, in the 2009 pandemic, a vaccine was developed within a year of the outbreak (There is no such assurance that a vaccine will be developed as quickly for the 2019 coronavirus); and, finally, (3) we don’t know the contagiousness and lethality of the coronavirus yet — only research and time will settle that question. In the meantime, prudent levels of action (such as lockdowns) are well-advised, according to Dr. Fauci.

Yet, there will always be people suspicious of the news media and politicians anytime drastic measures such as lockdowns and stay-at-home orders become common, particularly when it has a dramatic effect on the economy. At times during this pandemic, I have included myself in that skeptical crowd.

But not when it comes to testing.

Broad-based, systematic testing is a critical element of any national effort to contain a highly contagious and deadly virus, according to health organizations like the CDC and the World Health Organization (WHO). A country can’t address a problem effectively if it doesn’t understand the breadth of the problem.

As former NASA Flight Director Gene Kranz might say: ”Let’s work the problem people. Let’s not make things worse by guessing.”

Broad-based, systematic testing for the coronavirus removes some of the guesswork out of policy making during a pandemic.

Unfortunately, coronavirus testing is not being consistently implemented by U.S. states, as of 30 March. As Figure 1 (below) shows, the states that have conducted the fewest coronavirus tests per 100,000 people tend to be states that Trump won in 2016 — red states like Oklahoma, South Carolina, Texas, Alabama, Mississippi, Kentucky, Georgia and, of course, Iowa.

Figure 1: State-level coronavirus testing per capita (Data source: The COVID Tracking Project)

Republicans will rightfully point out that many of the states doing the most testing are also red states: Louisiana, Utah, North Dakota, Alaska, South Dakota and Montana. All have tested, so far, more than the national average of 359 coronavirus tests per 100,000 people.

Before concluding that there is a genuine political bias in the application of coronavirus testing, I tested a simple regression model to explain levels of coronavirus testing at the state-level. Included in my model were the following control variables: (1) the number of people hospitalized in the state for the coronavirus, (2) percentage of tests coming back negative (a proxy for states with broader testing procedures that include people without symptoms), and (3) an indicator variable for states where large (>50) infection clusters occurred (NY, WA, LA).

Additionally, I tested numerous political variables, two of which were significant predictors of state-level testing: (1) Trump’s statewide share of the 2016 vote, and (2) an indicator variable for states with a Republican governor but a Democrat-controlled state legislature.

According to the model, the relative importance of each variable in explaining state-level coronavirus testing are as follows (in ranked order; numbers in parentheses indicated standardized coefficients and the direction of the relationship):

• States with large-cluster infections (+0.65)
• Number of coronavirus patients hospitalized (-0.56)
• Percent of tests coming back negative (0.33)
• Trump’s share of 2016 statewide vote (-0.29)
• A Republican governor in a Democrat-controlled state (0.21).

The finding that states with the most coronavirus hospitalizations tend to have lower rates of testing, all else equal, is not as surprising as it sounds. Its significance is driven by the fact that the states with the most hospitalizations tend to be bigger states, which will also tend to have lower per capita levels of testing given their large populations.

More intriguing is the finding that Trump states, all else equal, are testing for the coronavirus at lower per capital levels. For every 10 percent of Trump’s 2016 vote share, the state tests 41 fewer people (per 100,000 people), all else equal.

Likewise, in Democrat-controlled states and holding other factors constant, the presence of a Republican governor adds 148 tests per 100,000 people.

See, not all Republicans are bad.

_______________________________________________________________

This may be the most important statistic that has so far emerged during the coronavirus pandemic: Twenty-five percent of those infected with the coronavirus may be asymptomatic, according to the CDC. These are people who carry the virus, show none of its COVID-19 symptoms, yet may still be contagious.

Super-spreaders are a likely source of many cases of the coronavirus worldwide.

To my mind, the super-spreader is the scariest aspect of the coronavirus. Given this virus kills between 0.1 and 3 percent of those who have it, identifying asymptomatic coronavirus carriers — or, at least, the prevalence of these people in the general population — should be among the highest priorities for government agencies charged with controlling the coronavirus.

Without large-scale levels of testing, including the systematic testing of people without any COVID-19 symptoms, it is impossible to know the extent of the super-spreader problem.

The time is long over for Republican governors in red states to operate under the fiction that the coronavirus pandemic is a phantasm of the liberal (Trump-hating) media. The time is now to lock-down and test.

• K.R.K.

To Iowa Governor Kim Reynolds: If your “internal data” is not telling you to lock-down, stay-at-home and do far more testing of Iowans, you need new “internal data.”

Requests for the data used in this analysis can be sent to: kroeger98@yahoo.com (It is also available at www.COVIDtracking.org)

APPENDIX:

A Linear Model of State-level Coronavirus Testing

By Kent R. Kroeger (Source: NuQum.com, March 27, 2020)

Every so often, when events seem to demand, political scientists crawl out of their cubby-holes and lecture the American people on how no Republican president can be popular without some combination of luck, war, deception, and a chronically lazy American public.

Political scientists have had a dim view of Americans ever since Angus Campbell, Philip Converse, Warren Miller, and Donald Stokes concluded in their book, The American Voter (1960), that the typical American voter has little more than a superficial, periodic interest in politics, over-estimate their knowledge about politics, and generally are incapable of organizing their political attitudes in a coherent manner that would resemble anything intellectually sophisticated or ideological.

This view has served political scientists much the same way the Big Bang’s cosmic background noise reminds astrophysicists that everything they measure and explain shares the same origin story.

In fairness, the research has evolved beyond the dark conclusions found in The American Voter, with some political scientists even recognizing the average American voter is remarkably adept at choosing political candidates that best match their opinions and attitudes.

Nonetheless, it seems like every time an American president — particularly one of the GOP variety — surges in job approval or popularity, academics and the news media have to set us straight.

The latest example is Donald Trump’s job approval rise since the start of the coronavirus (COVID-19) pandemic. RealClearPolitics.com’s aggregate poll summary of Trump’s job approval shows this clearly (see Figure 1):

Figure 1: Trump Presidential Job Approval (RealClearPolitics.com poll aggregate)

Like the grasping pincers of an earwig, the gap in Trump’s approval and disapproval has dramatically narrowed since mid-March when it became apparent that the coronavirus was not just a news story about China and Europe, but something that was going to effect every American’s daily life.

_______________________________________________________________

Mortified at the sudden relative popularity of a man who most political scientists and journalists probably despise, the cognoscenti have gone into high gear trying to explain why Trump’s rise in popularity is fleeting and won’t matter in November.

Harvard professor of public policy Matthew Baum does his best to sooth the fears of anxious Democrats staring into the abyss as their party prepares to nominate a man recognizably on the downward slope of his cognitive peak:

“Improvements in public approval ratings are notoriously ephemeral — usually lasting only a month or so. The COVID-19 rally seems unlikely to last long enough to help Trump’s reelection effort.

Indeed, as the campaign season approaches, the bipartisan support he has thus far received will almost certainly diminish. It’s impossible to predict how the current crisis will unfold. But unless the pandemic and all its attendant consequences are successfully resolved, the most likely scenario is that Trump has now reached his approval ceiling. If so, it may be downhill from there.”

NBC News, the most anti-Trump news organization — and its hard to pick just one for that title — coughed up this “hot take” on the sudden swell in Trump job approval:

“…Trump’s approval ratings — an important gauge of his re-election chances a bit over seven months away — are staying strong, by some counts even reaching highs for his tenure. In part, that’s a sign that no matter how much criticism Democrats lob at Trump, and whatever amount of negative media coverage he endures, his base supporters are resolute and unflinching. But the surge in support that goes beyond his base is a reflection of a familiar phenomenon whereby presidents become more popular in times of crisis, and according to history, his surge is puny — and likely to be short-lived.”

Puny is NBC News’ level of curiosity about why Trump, whose job approval has been far underwater since the start of his administration and has experienced the least variation of any president in the modern polling era, should suddenly climb at a rate unseen during his presidency (except for, perhaps, during the Senate’s Kavanaugh hearings in Fall 2018 and, in early 2019, when the Mueller investigation exonerated Trump of any Russiagate-related conspiratorial acts).

Since hitting rock-bottom job approval in December 2017, Trump’s job approval has increased at a tortoise’s pace of 0.36 points per month.

In the midst of a pandemic, Trump’s job approval has gained three percentage points in barely half a month. A three percentage point rise may seem “puny” relative to George H. W. Bush’s 31 percentage point gain prior to the First Gulf War or George W. Bush’s 39 percentage point surge after 9/11, but it must be viewed in the current context.

The dynamics behind presidential job approval have fundamentally changed in the past 30 years as intensifying hyper-partisanship within the U.S. Congress, the news media, and the general public has become the fundamental force in American politics. As a result, most Americans get their political news from the news source most aligned with their personal political views; and, in turn, makes aggregate presidential job approval much harder to move.

MSNBC and CNN basically run 24/7 with the same story: Trump is a lousy president. Fox News counter-programs with breaking news that Trump is still winning. And this news broadcasting model works for all the cable news networks as it produces fairly stable and profitable audiences.

Unfortunately, it also produces an overabundance of ill-informed, narrow-minded citizens.

[I will blow an aneurysm the next time I hear another Democrat defend Joe Biden’s babbling donnybrook of verbal nonsense as a function of his childhood stutter. Stuttering is not evidence of a cognitive deficiency. What at an ugly stereotype Democrats nonetheless are now pushing out about stutterers.]

_________________________________________________________________

To deride Trump’s approval gain as a temporary “rally-around-the-flag” effect is to miss the forest for the trees.

The most impressive feature of Trump’s approval is its consistent rise since December 2017, despite a national news media — Fox News and The Wall Street Journal notwithstanding— that, in his first 100 days, gave Trump the most negative news coverage of any recent president and has since maintained that negative tone.

If the national news media hammered on you the way it does on Donald Trump, do you think you would be as popular among your casual friends and acquaintances? Most certainly not.

How could the New York-Washington, D.C. media axis allow Trump’s approval to rise as it has in the past two plus years? The answer is simple: They win either way. The Trump presidency is the national media’s golden goose. And should the media axis powers defeat Trump in November, they win then too, as they will put into power a chief executive who won’t challenge their credibility and privileged status. That is what they call a win-win.

________________________________________________________________

So how why has the coronavirus pandemic worked in Trump’s short-term favor?

In contrast to the pundits dismissing Trump’s recent approval rise, as a public opinion researcher I talk in-depth to people who describe themselves as “undecided” or “independent.” They aren’t reflexive Trump supporters, but they aren’t anti-Trump either. And they may represent little more than 10 percent of the voting populace. But they matter electorally, should they overwhelmingly move for or against Trump in November.

Like all of us, swing voters directly experience the consequences of the coronavirus pandemic, but they are less influenced by national media narratives. They aren’t looking to blame the Trump administration for this crisis; but, rather, are seeking comfort in these times.

In my research with swing voters, Trump’s daily news conferences on the coronavirus pandemic has given them a sense of optimism, in contrast to the incessant gloominess of the national media.

The most recent World Health Organization (WHO) data on the coronavirus epidemic supports at least a modest level of optimism (see Figure 2). Italy and Spain, the two hardest hit European countries, are discernibly past the peak growth of the virus — both experiencing a decline in new daily cases about 20 days after the initial surge in coronavirus cases.

Figure 2: New Daily COVID-19 Cases for Selected Countries

This empirical evidence supports Trump’s optimism that the U.S. will get past the worst of the coronavirus pandemic by mid-April. Forecasts by the University of Washington’s Institute for Health Metrics and Evaluation (IHME) further buttress Trump’s confidence when they project that the number of deaths-per-day and the stress on hospital resources will peak around April 15th.

Still, according to the IHME, the U.S could see 80,000 deaths due to COVID-19, a significant increase over the usual number of communicable disease related deaths the U.S. normally experiences annually.

Should the Trump administration bear some responsibility for the apparent inadequacy of its initial response to the coronavirus pandemic? Yes, of course — just as most European countries bear some responsibility for the same failure.

________________________________________________________________

There is a reason why MSNBC is not covering the coronavirus pandemic to the same extent as Fox News, and even CNN. They realize it helps Donald Trump (see Figure 3).

Figure 3: Percentage of News Network Airtime Devoted to the Coronavirus by MSNBC, CNN and Fox News (Source: The GDELT Project)

The coronavirus pandemic is humanizing Trump. His proclivity for bombast and unearned confidence actually serves the public good in these times.

People want optimism when they are feeling uncertain and stressed. Trump oozes such sentiments; in contrast, the Democrats wallow in gloom and doom.

This pandemic is going to loom over the general election in November, whether or not it is completely vanquished by then. In this time, Trump holds a tremendous advantage over the Democrats. He is the Optimist-in-Chief.

________________________________________________________________

Donald Trump is the tortoise and the Democrats are the hare in the Aesop’s Fables classic story. Trump is slowly building support among swing voters and disenchanted Democrats, most recently through his very human attempt to sooth our fears in the midst of the coronavirus pandemic. The Democrats, meanwhile, revealed their indifference to average Americans by voting unanimously with the Republicans for the Coronavirus bailout program, the biggest wealth transfer to wealthy Americans in history.

Two (only marginally different) Republican parties dominate the U.S. political system. The Democrats’ Vote Blue No Matter Who mantra doesn’t sell to intellectually honest people.

Yes, Trump is not a good person, but he is a transparent one. He shows us what he is everyday and doesn’t hide it.

I can’t say the same for most Democrats.

• K.R.K.

Offended Democrats can send their complaints to: hello@tomperez.org

By Kent R. Kroeger (Source: NuQum.com, March 26, 2020)

Nobody truly knows when the coronavirus pandemic will end and its final human toll.

That doesn’t stop some of us from trying to predict how this crisis will resolve.

I predicted in late January that the coronavirus (2019-nCoV) — the virus that causes the COVID-19 disease — would peak in China in mid February around 20,000 cases, effectively joining the pantheon of pandemic scares that never materialized in a big way on a global scale (e.g., 1976 Swine Flu, 2002 SARS, 2012-present MERS, etc.).

After that bad call, I’m officially out of the coronavirus prediction business.

I know I don’t know what is ultimately going to happen with COVID-19.

But the real experts don’t know either.

Most of us have visited the Johns Hopkins University (CSSE) website that provides “near real-time” counts of coronavirus COVID-19 cases and deaths worldwide.

Their website has numbers — lots of them — from reliable government sources, nonetheless. Those numbers must be fairly accurate, to the extent possible under such difficult circumstances, right?

Well, no. We don’t know the accuracy of the WHO/JHU(CSSE) numbers are yet.

Scientists who think most about viruses and pandemics are only now beginning the process of understanding the true numbers surrounding this pandemic and how it will end — and we know it will end, someday.

When viewed across countries, the numbers reported by the World Health Organization (WHO) and the JHU (CSSE) website should be taken with a grain of salt. These numbers are probably significant under-counts of the true numbers of COVID-19 cases and related-deaths — numbers that we will know with some confidence perhaps many months (or even years) from now.

At some point, probability samples within those countries most affected by the coronavirus will reveal the percentage of their populations with the virus’ antibodies, indicating how many people were actually infected.

Those types of epidemiological studies take time and aren’t amendable to the 24-hour news cycle that drives our current dialogue around the coronavirus.

Yet, we need good information about the coronavirus now, not eight months from now. Effective public policy requires as accurate, timely information. How contagious is the coronavirus? What is its mortality rate? How does its mortality rate vary by demographic categories (i.e., who is most vulnerable)? What percentage of those infected with the virus failed to ever show symptoms?

________________________________________________________________

The problem is that there is no single simulation model for assessing the trajectory of a pathogen like the coronavirus. Different models, requiring a different sets of assumptions, can render different conclusions. May the best model win.

Presently, a disagreement between two academic groups in the United Kingdom studying the coronavirus highlights this analytic mélange.

In one corner are researchers from Imperial College (London, UK) who, after exhaustive research on individual-level coronavirus transmission dynamics, ran simulations on how the coronavirus would put pressure on the Great Britain (GB) and US health care systems depending on the non-pharmaceutical interventions (NPIs) adopted by the two countries.

Should either country fail to adopt any control measures, according to the Imperial College study, GB and the US can “expect a peak in mortality (daily deaths) to occur after approximately 3 month.”

“In such scenarios, given an estimated R0 of 2.4, we predict 81 percent of the GB and US populations would be infected over the course of the epidemic,” the study concludes. “For an uncontrolled epidemic, we predict critical care bed capacity would be exceeded as early as the second week in April, with an eventual peak in ICU or critical care bed demand that is over 30 times greater than the maximum supply in both countries.”

[R0 represents the expected number of infection cases directly generated by one infected person in a population where all individuals are susceptible to infection.]

That’s the worst-case scenario.

In contrast, the study finds that the most effective combination of interventions is predicted to be “a combination of case isolation, home quarantine and social distancing of those most at risk (the over 70s).”

The Imperial College study is credited with changing GB policy on the coronavirus almost immediately upon its release.

In another corner, researchers from Oxford University (UK), employing a different simulation method, come to a drastically different conclusion about the coronavirus.

“The results we present (in this study) suggest the ongoing epidemics in the UK and Italy started at least a month before the first reported death and have already led to the accumulation of significant levels of herd immunity in both countries,” write the authors of this, as yet, unpublished study.

Herd immunity is critical in dampening or stopping the spread of a virus like the coronavirus. The higher the population percentage already carrying the virus antibodies, the less likely the virus can spread within those still vulnerable to it.

The Oxford study suggests that the overwhelming majority of those infected by the coronavirus never become sick enough to need hospital care. The public policy implications are profound: The coronavirus was spreading rapidly within the population at least a month before the first case was confirmed. In other words, it may be too late for any suppression and mitigation strategies to have a major impact on the trajectory of the coronavirus. The cake is already baked, as they say.

Despite the study’s key findings, the study’s lead researcher Professor Sunetra Gupta, still believes behavioral changes, such as social distancing and self-quarantines, will still be effective at reducing the number of people who become seriously ill due to the virus.

Even if the Oxford study is found to be substantively wrong, Professor Gupta understands the magnitude of the research task needed to render that conclusion: “We need immediately to begin large-scale serological surveys — antibody testing — to assess what stage of the epidemic we are in now.”

[I should note that the Oxford study has come under intense criticism from some prominent epidemiologists since it was released to the media. A sense of that criticism can be found here.]

________________________________________________________________

It is popular in some circles now to dismiss the coronavirus pandemic as more media hype than a globe-threatening health crisis. And, superficially, the numbers support that cynicism.

As of 8:15am today (March 27, 2020), the coronavirus has killed a reported 24,361 people worldwide. There have been five other pandemics in the past 125 years with death tolls far greater than the 2019 coronavirus.

Figure 1: Previous pandemics (Name, Time Period, Source, Death Toll)

HIV/AIDS has killed over 25 million people since its first case in 1981. That is about 625,000 people a year. The 2009–10 Swine Flu killed 200,000 people worldwide (12,000 in the U.S.). The first pandemic I vaguely recall — the 1968–70 Hong Kong Flu — killed 1 million people worldwide, including 100,000 in the U.S.

I don’t recall large chunks of the U.S. going into a stay-at-home lockdown in 2009 (…because we didn’t). There weren’t daily presidential news conferences about the Hong Kong Flu in 1969.

Why is the coronavirus being treated differently?

Keeping in mind that the statistics for the new coronavirus (COVID19/2019-nCoV) are moving targets (for example, estimates on its mortality rate range from 1 to 3 percent), the panic over this latest pandemic most likely stems from the combination of its contagiousness and mortality rate.

The 2014–16 (West African) Ebolavirus killed almost half of the 29,000 people that contracted the virus. It had a high mortality rate, but was geographically concentrated. While it was highly contagious, its symptoms were quite visible and the ability to identify and isolate its carriers much easier than the coronavirus.

The two other coronavirus outbreaks (SARS, MERS) were qualitatively different from the current coronavirus outbreak as well. While SARS had a mortality rate around 10 percent, it was less contagious than 2019-nCoV and, in the end, only infected 8,500 people (mostly in China).

The ongoing MERS outbreak has killed 912 people out of the 2,494 who have contracted the virus (a more than 30% mortality rate). But, again, MERS has been geographically isolated and does not appear to be nearly as contagious as 2019-nCoV.

The 2009 Swine Flu is perhaps the most instructive comparison to 2019-nCoV. Like the current coronavirus, the 2009 pandemic received considerable media attention when the WHO announced the potential for a global pandemic from this new H1N1 flu virus in April 2009. However, in the end, the 2009 Swine Flu’s mortality rate of roughly 0.1 percent was not substantially different from the common flu; and, while 12,000 people did die in the U.S., perceptually it felt like just a “bad” flu season.

The new coronavirus (2019-nCoV) is fundamentally different from these other pandemics. Where the common flu typically has a mortality rate around 0.1 percent (1 in every 1,000 cases), the current coronavirus most likely has a mortality rate at least 10 times more lethal.

Lastly, the numbers commonly cited for pandemics such as the 2009 Swine Flu are epidemiological estimates derived after those outbreaks have ended. Epidemiologists go back into the aggregate diagnostic and mortality data and look for evidence of what impact the pandemic virus may have actually had (beyond simply what was reported contemporaneously). They make sophisticated estimates with known (and unknown) amounts of error. \

The numbers for COVID-19 are reported to the WHO by national health ministries. At a minimum, we know by anecdote that not everyone who contracts a virus goes to the hospital. Many (maybe most) resort to home-care and never seek medical attention. Some will contract a virus and never exhibit symptoms. Evidence of this latter fact comes from the NBA where hundreds of players and coaches exposed to carriers of the coronavirus were tested.

“Following two players testing positive last week, others were tested and five additional players tested positive,” the NBA issued in a statement last week. Since March 11, when Utah Jazz player Rudy Gobert was the first NBA to be diagnosed with the coronavirus, thirteen other NBA players have tested positive. From the most recent information made available to the public, not one of these players has experienced significant symptoms from the coronavirus — which reinforces what may be the most significant feature of the new coronavirus: most infected by it will have no (or only minor) symptoms related to it.

The occurrence of asymptomatic coronavirus carriers has also been revealed by the Germans who have done significantly more coronavirus testing (per capita) than the U.S.

As of March 19, the U.S. had conducted in total only 104,000 tests, according to The COVID Tracking Project. In contrast, according to German health officials, they are equipped to conduct 160,000 tests per week and, unlike the U.S., systematically test people who are not symptomatic. Germany’s broader testing effort is designed to understand the true extent of the coronavirus infection rate and has revealed thousands of Germans who carry the virus but are not symptomatic.

As of March 27, the Germans have reported 304 deaths against 49,344 confirmed cases, for a mortality rate of 0.6 percent — among the lowest rates in those countries most affected by the coronavirus. Some epidemiologists are citing Germany’s effort to identify asymptomatic carriers early as one of the reasons for Germany’s low mortality rate. Identify those infected before they are sick and you increase their chance of a full recovery.

But more than asymptomatic carriers and those who treat their symptoms exclusively at home can fail to get counted by health officials. Research has shown that medical databases are not always accurate. For example, people can die from complications associated with a virus (organ failure, hemorrhaging, etc.) and yet never get classified as having died from the virus if they had not been tested for it.

The reality is that data collection on a scale as large as the coronavirus requires is going to have significant amounts of error through selection bias, incomplete population coverage, coding mistakes and simple random chance.

I will make this one prediction about COVID19: The numbers were are seeing now (as reported to the WHO) are nowhere near the true impact of COVID-19 — particularly in countries with less-developed health care systems and limited resources for health testing and monitoring.

When this is all over, it will take many months before we know what COVID-19 really did.

__________________________________________________________________

Even so, I believe there is evidence in the current COVID-19 data from the WHO that this pandemic will peak soon.

We have already seen countries get control of their first wave of the coronavirus — China, South Korea — and Italy, U.S. and France may be making this critical turn soon (see Figure 2).

Figure 1: Coronavirus trajectories of daily new cases for selected countries

The current coronavirus pandemic will end. It is not a question of if, but one of when.

But I’m not making any predictions.

• K.R.K.

Datasets used in this article are available by request to: kroeger98@yahoo.com

By Kent R. Kroeger (Source: NuQum.com, March 23, 2020)

Health experts continue to warn us that the confirmed coronavirus cases reported by the World Health Organization (WHO) and national health ministries are significantly under-counting the actual number of cases in the U.S. and worldwide.

Earlier this month Dr. Marty Makary, a surgeon and health policy expert at Johns Hopkins University, told Yahoo Finance’s “On the Move” that when the U.S. Centers for Disease Control and Prevention (CDC) was reporting 1,600 confirmed cases in the first week of March, the actual number of infected people could be “between 50,000 and half a million.”

“Don’t believe the numbers when you see, even on our Johns Hopkins website, that 1,600 Americans have the virus,” Makary said. “No, that means 1,600 got the test, tested positive. There are probably 25 to 50 people who have the virus for every one person who is confirmed.”

Dr. Makary’s estimated range of coronavirus cases in the U.S. mirrors my own cross-sectional analyses of the WHO and Johns Hopkins databases.

As of March 21st, I estimate that the actual number of Americans infected with the coronavirus (including those who have died or recovered) numbered between 62,000 and 215,000.

[The methodology I employed for these cumulative case estimates is described here.]

Building upon a country-level coronavirus case estimation model I developed in an earlier article, I added a control variable (NUM_TEST) that measures the number of coronavirus tests conducted by each country through March 18th (53 countries). This variable is being collected ourworldindata.org and, as of today, does not have reliable testing data for a number of important countries, including China, Canada and most South American and African countries.

Writes ourworldindata.org about this coronavirus testing variable:

“To understand the statistics on cases, we would want to know how many people in the world are being tested for COVID-19 every day, and how the available tests are being allocated. Unfortunately there is no centralized WHO data on COVID-19 testing and most countries do not provide official reports on tests performed.

Some countries do provide aggregate estimates on the total number of tests performed up to a specific date, but these are published across individual websites, statistical reports and press releases — often in multiple languages and updated with different periodicity.

Given the large problems with data availability, we did a manual review of data across national reports, and included the most recent estimates that we could find as of 20 March 2020, 18:00 GMT.”

Given the uncertainties with how countries are reporting their coronavirus testing efforts, I report two estimation models in this article: Model 1 (which does not include a control for the number of coronavirus tests) and Model 2 (which does include a control for the number of coronavirus tests).

The Results

With each new day that the WHO reports confirmed coronavirus cases, my original model (Model 1) has increased its ability to explain country-level variation in coronavirus cases. The original March 14th model explained 73 percent of the variance in country-level case numbers and found the following variables to be significant predictors: National income per capita, total population, number of international tourism arrivals per year, percent of annual deaths due to communicable diseases, and an indicator variable for China.

Estimating Model 1 using WHO data through March 21st found that the model now explains 81 percent of the variance in country-level case numbers (see Appendix below for the regression output for Model 1).

An analysis of the residuals for Model 1 found the model significantly over-estimated the actual number of coronavirus cases in the U.S., Japan, Myanmar, New Zealand, and the United Arab Emirates (see Figure 1).

Conversely, the model under-estimated the actual number of cases found in Iran, Italy, and South Korea.

Figure 1: Residual analysis (Actual vs. Predicted) for Model 1 (data through 21 March 2020; 167 countries)

Why are we seeing these significant over- and under-estimates?

Using the U.S. as an example, Model 1 estimates that the U.S. — given its national wealth, population size, size of its tourism industry, and experience with communicable diseases — should have reported 215,000 coronavirus cases on March 21st. Instead, the U.S. was reporting 25,000 cumulative cases on that date.

Why the big difference?

There can only be one of two answers: (1) The U.S. has made an heroic effort to keep the coronavirus from spreading (e.g., widespread testing, social distancing, self-quarantines, travel bans, local and statewide lockdowns, etc.), or (2) the U.S. has systematically under-counted the actual number of coronavirus cases in the country and is now playing catch-up with its ramped up testing regimen.

I suspect the smart money is on the latter explanation.

But that explanation may not work for a country like UAE — which reported 154 cases (as of March 21st), though Model 1 expected 1,250 cases.

According to ourworldindata.org, the U.S. has conducted 103,945 coronavirus tests as of March 19th (or about 32 tests for every 100,000 people). Contrast the U.S. number to UAE where there has been 1,298 tests per 100,000 people.

Other countries with a high number of coronavirus tests relative to their population size include: Australia (454 per 100,000), Bahrain (1,188 per 100,000), Norway (823 per 100,000), and South Korea (613 per 100,000).

One possible hypothesis: Whereas the U.S. may have missed the opportunity to prevent a large-scale spread of the virus and is now playing catch-up with its recent increases in testing, the UAE, through its aggressive testing regimen early in the outbreak, may have stunted the spread of the virus to the point where the apparent under-count (according to Model 1) is actually the true value.

In other words, Model 1 isn’t properly specified— we need to add a variable measuring a country’s coronavirus testing efforts.

Model 2, therefore, includes the ourworldindata.org collected data for the number of coronavirus tests at the country-level (Model 2 regression output can be found in the Appendix).

Regrettably, inclusion of the number of coronavirus tests variable eliminated over 100 countries — mostly from Southeast Asia, Africa and South/Central America — from the analysis in Model 2. Nonetheless, Model 2’s parameter estimates were fairly consistent with the Model 1 results (see Appendix for model output and diagnostics), though the model fit fell substantially (from 81 percent of the variance explained, down to 61 percent).

Given the loss of data from the southern hemisphere (and China), the Model 1 variable for communicable disease deaths and the indicator variable for China were dropped.

The remaining variables in Model 2 were significant (numbers in parentheses are standardized coefficients indicating their relative importance): National per capita income ( 0.475), population size (0.240), annual number of tourism arrivals (0.344), and cumulative number of coronavirus tests (0.257).

Figure 2 shows that the outlier nations in Model 2 and the direction of that error. Superficially, Model 2 outlier nations look similar to Model 1. The predicated values for U.S., New Zealand, and UAE remain higher than actual numbers; while Iran, Spain and Italy are experiencing numbers higher than expected based upon their population size, wealth, and annual tourism volume.

Figure 2: Residual analysis (Actual vs. Predicted) for Model 2 (data through 21 March 2020; 53 countries)

For now, at least, we can’t explain UAE’s lower-than-expected count of coronavirus cases to their aggressive testing program. Something else explains their low reported numbers.

Instead, Model 2 offers a glimpse into how the WHO reported numbers differ from expectations based on established covariates with the spread of the coronavirus:

Figure 3: Model 2 predictions versus reported coronavirus case numbers

As of March 21st, at a minimum, the U.S. has almost 62,000 cumulative cases of the coronavirus instead of the 25,468 reported to the WHO.

At a maximum, according to the more inclusive Model 1, the U.S. has over 215,000 cumulative cases of the coronavirus.

Either way, the U.S. is seriously under-estimating the extent of the spread of the coronavirus within its borders.

• K.R.K.

Datasets and statistical software used in this article are available upon request to: kroeger98@yahoo.com

APPENDIX:

Variables

Dependent Variable (Source: WHO): Number of confirmed COVID-19 cases as of 21 March 2020 (log transformed)

Independent Variables (Source: World Bank):

NY_ADJ_NNTY_PC_CD = Adjusted net nat’l income per capita (current US$)

LOG_POP = Country’s total populaton (log transformed)

CHINA = Indicator variable (1 = China; 0 = all other countries)

SI_INT_ARVL = International tourism, number of arrivals

SH_DTH_COMM_ZS = Cause of death, by communicable diseases and maternal, prenatal and nutrition conditions (% of all deaths)

NUM_TESTS = The number of coronavirus tests reported through March 18th by each country’s health ministry or national authority (Data provided by ourworldindata.org and can be found here)

More detail on these independent variables (except NUM_TESTS) can be found in the World Bank’s data catalog at: https://datacatalog.worldbank.org/dataset/world-development-indicators

Linear regression models

Model 1 (Does not include a control for the number of coronavirus tests; 167 cases)

Model 2 (Includes a control for the number of coronavirus tests; 53 cases)

Residual plots

Model 1 (Does not include a control for the number of coronavirus tests; 167 cases)

Model 2 (Includes a control for the number of coronavirus tests; 53 cases)

By Kent R. Kroeger (Source: NuQum.com; March 21, 2020)

For three years the U.S. news media wasted our time promulgating a conspiracy theory that candidate Donald Trump and his campaign conspired with the Russians to steal the 2016 presidential election.

It was never true. Not even a little bit true. It was a conspiracy theory built on half-truths, coincidences, profit motives and malicious, partisan intentions.

And now, with the coronavirus pandemic, we are witnessing a new bull market for conspiracy theories:

• In the midst of a contentious trade war with China, the U.S. military planted the coronavirus in China to weaken their position,
• The Chinese military used the coronavrius outbreak to end the democracy protests in Hong Kong,
• The coronavirus was released due to a lab accident in Wuhan, China,
• The pharmaceutical companies created the coronavirus knowing they already had the drugs necessary to treat it (thereby enriching themselves),

There are many more conspiracy theories about the coronavirus. Too many. All predicated on nothing more than circumstantial evidence — mere coincidences — and implanted into the world’s information bloodstream where they prosper, morph, and instill a pernicious type of distrust in a vulnerable segment the general population desperately looking for explanations for why they are now terrified to leave their homes.

Perhaps actress Kirstie Alley describes many of us are attracted to conspiracy theories regarding the coronavirus (no matter absurd their internal logic):

Many of us have a preternatural need to explain events and phenomena that shake our security and sense of stability.

Sometimes we just can’t accept that random coincidences happen without some larger purpose or meaning. Who wants to live under the constant threat of random chance?

_________________________________________________________________

One of the first challenges greenhorn intelligence analysts face in the U.S. Intelligence Community — as they analyze data and information related to their area of expertise —is learning how to distinguish random coincidence from significance.

Two or more events can appear through their ordering or simultaneity to be significantly related — when, in fact, they are not.

We’ve all said this at one point in our lives: What are the chances?

Two years ago, on an unremarkable Saturday morning, I realized I had not done laundry for a while. I despise doing laundry. In my 50s now, I have never separated whites from colors during the washing process. Never. Not once.

On that day, I couldn’t find even one clean shirt to throw on before getting my morning coffee and donut at the local supermarket. In desperation, I pulled a garbage bag full of old shirts out of my closet that I had never unpacked since moving from Iowa to New Jersey three years ago.

I knew if I dug down too deep in the bag, grody smells might escape, so I just yanked out what was on top: a crisp white t-shirt I had purchased in commemoration of Chinese leader Xi Jinping’s February 2012 visit to Iowa.

The t-shirt had never been worn. It was meant to be a souvenir. But, in a desperate moment, I threw it on — along with some jeans — and drove to the market.

While waiting for my double-shot espresso, I decided to pick up some recliner chair food for watching the day’s schedule of college basketball: some deli meats, sliced cheese, fresh bread and Nabisco animal crackers. As I took a tight turn into the cracker aisle, I was forced to make a Euro step move to avoid a man walking out of the same aisle.

We apologized to each other and instantly realized something bizarre had just happened: we were both wearing a Xi Jingping (Welcome Back to Iowa) t-shirt.

What are the chances?!

We both spit out a laugh and chatted long enough to discover we had lived in Des Moines around the same time and were professionally interested in China.

But what were the chances that we’d wear those identical t-shirts on the same morning, in the same supermarket, in the cracker aisle, in Pennington, New Jersey?

No need to do the math. The probability was not substantively different from zero.

Yet it happened. It was a random coincidence — nothing more significant than that.

________________________________________________________________

Other coincidences are not as improbable, even if we think they are.

The classic case, often used in introductory college statistics courses, is the Birthday Problem: What is the probability that at least two students in a classroom of 30 share the same day of birth?

The typical first-year statistics student thinks the chances are slim; in my experience, they will guess something well under 50 percent. There are 365 days in a year, after all. Two randomly selected people have a 0.27 percent chance of sharing a day of birth. How much better are the chances in a room of 30 students?

It turns out, a lot. Throwing a simple probability formula against the problem, the chances of at least two students in a class of 30 sharing the same day of birth is 71 percent.

The lesson in these two coincidence examples is that people are prone to making judgment errors in presence of natural randomness.

The Xi Jinping t-shirt story is an illustration of Littlewood’s Law which states that a person can expect to experience events with an odds of one-in-a-million at the rate of one per month.

It’s not a formal scientific law, but it brings attention to the omnipresence of improbability in our daily lives. The improbable is everywhere.

For example, what are the chances that as I write this essay wearing only one sock, my son cries out about not being able to find his science class notebook, at the same time my dog starts barking at the deer in our backyard, at the precise moment my wife walks into the kitchen and complains that I didn’t wash the dishes from last night’s dinner in which I used a new Indian curry recipe she was given while in India last January?

In combination with how generally poor we are at assigning accurate probabilities to even common events — an intrinsic cognitive flaw in how our brains are wired — natural randomness can bring significant stress to our lives.

________________________________________________________________

Science writer Michael Shermer calls this cognitive flaw “folk numeracy,” which he defines as “our natural tendency to misperceive and miscalculate probabilities, to think anecdotally instead of statistically, and to focus on and remember short-term trends and small-number runs.”

As a result, we frequently exaggerate the probabilities of some threats (e.g., being a victim of terrorism) and wrongly minimize others (e.g., getting in a car accident while we are driving).

In his book, How We Believe, Shermer describes the human brain as evolving out of necessity into a pattern recognition machine that creates structure and meaning so we can navigate our surroundings with minimal stress. Our brain is an efficient uncertainty reducer that aids us in survival and reproduction, says Shermer.

Psychologist R. Nicholas Carleton, an expert in anxiety disorders, ties the human need to recognize patterns to a deep-seated fear-of-the-unknown — what he calls, our fundamental fear.

Not only is fear-of-the-unknown inherent to the human species, according to Dr. Carleton, it is “distributed and normally in the population.” Some people have more of that fear than others — and some less. As such, someone with a strong fear-of-the-unknown but relatively poor at bringing order to their world (e.g., assigning accurate probabilities to events) will potentially suffer high levels of anxiety.

________________________________________________________________

Coincidences are manifestations of chance and our strategies for interpreting these improbabilities can vary significantly.

Most often, the statistician in me sees the joint occurrence of improbable events (coincidence) as unremarkable. Unless you bring a defensible theory and independent data to the explanation, don’t waste my time.

Yet, there are times when coincidences do reveal something more causally significant. As mystery writer Simon Van Booy once put it, “Coincidences often mean you’re on the right path.”

The current controversy swirling around Tennessee Senator Robert Burr comes to mind. According to a March 19 ProPublica story, “soon after he offered public assurances that the government was ready to battle the coronavirus, the powerful chairman of the Senate Intelligence Committee, Richard Burr, sold off a significant percentage of his stocks, unloading between $628,000 and $1.72 million of his holdings on Feb. 13 in 33 separate transactions.”

Days later, worldwide stocks markets dropped precipitously. Now that is some coincidence. Actions by at least one other U.S. politician is under similar suspicion.

Time will tell if Senator Burr’s actions were criminal, but the story highlights an example where coincidence cannot be dismissed out-of-hand as random chance.

Occasionally, even if rarely, conspiracy theories are right. The trick is knowing when.

• K.R.K.

Send comments to: kroeger98@yahoo.com

By Kent R. Kroeger (NuQum.com, March 16, 2020)

Two news stories this morning (March 16, 2020) caught my attention.

The first story concerned Utah Jazz basketball player Donovan Mitchell who was diagnosed to have the coronavirus (2019-nCoV) and apparently became infected through his close contact with fellow player Rudy Gobert, who also tested positive for the virus.

“I’m asymptomatic,” Mitchell told ABC’s “Good Morning America” on Monday. “I don’t have any symptoms. I could walk down the street (and) if it wasn’t public knowledge that I was sick, you wouldn’t know it. I think that’s the scariest part about this virus. You may seem fine, be fine. And you never know who you may be talking to, who they’re going home to.”

We should assume there are thousands of Donovan Mitchell’s roaming our country feeling healthy and promoting the spread of COVID-19.

The second story I heard on public radio station WNYC this morning. While being interviewed on the The Brian Lehrer Show, Dr. Irwin Redlener, professor of pediatrics and director of the National Center for Disaster Preparedness at Columbia University, told the NPR audience that the emphasis for containing COVID-19 must focus on those already sick and not trying to test as many people as possible for the virus.

I am not a medical doctor or epidemiologist, but as a statistician, those two stories raise a number of red flags.

First, addressing Dr. Redlener’s comments, to suggest our nation’s efforts must be focused on those already sick is a mea culpa suggesting our hope of limiting the spread of this virus is a misuse of limited resources.

According to many health care experts, our health care infrastructure is under too much stress taking care of those infected with 2019-nCoV to take on the responsibility of measuring its incidence.

If he is right, trying to measure the prevalence of COVID-19 in the population is purely an academic exercise that distracts from saving lives.

But then came the Donovan Mitchell story. Infectious disease experts and epidemiologists readily admit they don’t know the percentage of COVID-19 carriers are asymptomatic (i.e., they show no symptoms of the virus while they are contagious).

How could they know? This is a new virus. They call it ‘novel’ for reason.

It seems reasonable to assume the number Americans currently infected with the coronavirus is greater than the 4,464 confirmed cases reported to the WHO (as of 16 March).

_________________________________________________________________

In a previous essay I estimated there were almost 20,000 “unmeasured” Americans infected with the coronavirus (as of 14 March). However, my effort was exploratory, not conclusive, if only because I assume the reported U.S. numbers up to now are an underestimate caused by a flawed government response to the virus. However, an equally valid conclusion could be that the relative low number of U.S. confirmed cases is the result of successful state and federal government efforts to contain the virus.

Frankly, how can we single out the Trump administration when so many European and economically-advanced countries have made the same mistakes? If anything, this crisis exposes the inadequacy of the advanced economies, in general, to manage global pandemics.

One exception is South Korea which saw its number of new confirmed cases start to decline over a week ago (see the orange line in the chart below).

According to experts, South Korea’s aggressive testing effort deserves credit for this coronavirus success story. From the moment the coronavirus starting appearing in South Korea the country has fielded almost 190,000 tests — approximately 3,690 tests per 1 million people.

In contrast, as of 8 March, the U.S. has conducted only 8,400 tests resulting in 213 positive cases (keep in mind, some patients have been tested more than once). If my conclusion is correct that 19,000 Americans currently have coronavirus but have not been tested, the U.S. will need 790,000 tests kits just to test these “unmeasured” people. But even that number is an underestimate as it does not include the Americans who are still to get COVID-19.

Conservatively, if the test’s 2.5 percent yield holds, the U.S. needs millions of test kits to identify the majority of Americans infected by the coronavirus. In that light, Florida Governor Ron DeSantis’ recent announcement that his state will be distributing 500 more test kits to hospitals and private labs feels grossly inadequate.

Even so, Dr. Redlener’s concern has merit that too much emphasis on testing will take resources from hospitals and doctors tasked with treating those already sick. Health care professionals are already reporting shortages due to this pandemic.

Therefore, should the U.S. even try to diagnose every coronavirus carrier?

_________________________________________________________________

Dr. Redlener’s answer appears to be ‘No.’ And he may be right, but my question is: Why should the wealthiest country in the world need to choose between treating its citizens for a deadly virus and accurately measuring the prevalence of this virus in the general population?

I believe its a trade-off we don’t need to make and will only cost us more in the long run as “unmeasured” coronavirus carriers will extend this outbreak even farther into the future.

“Asymptomatic and mildly symptomatic transmission are a major factor in transmission for Covid-19,” Dr. William Schaffner, a professor at Vanderbilt University School of Medicine, told CNN. “They’re going to be the drivers of spread in the community.”

I was reminded of Dr. Schaffner’s warning when I read the story of Nebraska basketball coach Fred Hoiberg being taken to the hospital after he showed flu-like symptoms during his team’s Big 10 basketball tournament game against Indiana. Hospitalized after the game, it was determined Coach Hoiberg had the normal flu (Influenza-A) and not COVID-19. All’s well that ends well, right?

No, not right.

Coach Hoiberg didn’t fall ill during the game, he felt ill before the game. In fact, he felt bad enough to visit the Big 10 tournament doctor an hour before the game started. And what did the tournament doctor conclude? Despite showing flu-like symptoms — in the middle of the coronavirus crisis — the doctor decided Hoiberg was OK to coach from the bench.

If Hoiberg and that Big 10 doctor still have jobs today, the University of Nebraska and the Big 10 should be ashamed of themselves.

Being a self-obsessed college basketball coach is not an excuse for being so dismally unaware of the world around you that you think its OK to potentially expose hundreds of players and fans to a pandemic virus. What was Hoiberg thinking?

His words:

“Please let it be known that I would never do anything that would put my team, family or anyone else in harm’s way. I was feeling under the weather on Wednesday and we felt the right thing to do was to get checked by a tournament doctor prior to our game in the Big Ten Tournament against Indiana. Once that medical official cleared me, I made a decision to coach my team. I would like to thank event staff for their care and professionalism. Also, thank you to everyone who has reached out for your support. This is a scary time for all of us. Let’s offer our thoughts and prayers directly to those affected with Coronavirus.”

Hoiberg proves you don’t have to be smart to be a NCAA Division I basketball coach, but the tournament doctor? What was his excuse for letting Coach Hoiberg into the arena? There is no possible way this doctor could have tested Hoiberg for the coronavirus by examining him right before the game’s tip-off. Clearly, the doctor was more afraid of affecting the outcome of a Big 10 basketball game between two of the conference’s worst teams than he was about potentially spreading a deadly virus to hundreds of people.

This Big 10 tournament doctor needs to appear before a medical review board to defend why he still has a medical license.

________________________________________________________________

If my March 14th analysis is accurate, there may be nearly 20,000 Americans walking around with the coronavirus but undiagnosed. How do we find them?

From a statistical sampling perspective, the task is daunting. According to my analysis on 14 March, the incidence of undiagnosed coronavirus carriers was 20,000 ÷ 327 million U.S. citizens (or about 1-in-16,000).

Any effort to find these 20,000 people is helped by the fact that these cases will tend to be geographically clustered. We already see this clustering among diagnosed coronavirus cases in the U.S. (e.g., New Rochelle, NY; Biogen conference in western Massachusetts, Life Care Center of Kirkland, near Seattle, WA). This clustering is a function of the coronavirus’ R0— an estimate of the number of people an average infected person spreads the virus to — which the current evidence indicates is between two and three, making it substantially more contagious than the typical flu.

For sampling purposes, a higher R0 assists in detection. Find one infected person and you are likely to find many more.

A back-of-the-envelope estimate of the testing sample sizes needed to detect and isolate areas containing infected persons goes as follows:

First, let us assume the probability distribution of those infected follows a binomial form:

where f(x) is the probability x is the number of undetected infected persons in a sample of size of n. The value of p is the hypothesized proportion of undetected infected persons in the population.

From the above equation we can begin to estimate the sample sizes needed to find coronavirus carriers.

Let us start with a simple case: Assume you live in a city of 16,000 people. Based on our hypothesized incidence rate of 1-in-16,000, there may be one person in your city with the coronavirus. How do we find this person? We test a random sample of the city population, but since tests are expensive to administer, we want to minimize the size of our sample. Therefore, how big does the sample need to be to have a reasonable chance of detecting this one infected person?

Acceptance sampling methods allow us to find a happy middle ground between a 100-percent-census (which is cost prohibitive) and no measurement at all.

Since we consider a population “infected” (and, presumably, quarantined) if we find even one infected person, acceptance sampling (at a General II inspection level) suggests we need a sample size of 315 randomly selected people.

Scaling up to a larger population, a city of 100,000 would be stratified into 6 distinct geographic areas with approximately 16,000 people in each geographic strata. Again, we’d take a sample of 315 people in each strata for a total sample of 1,890 people.

Scaling up further to the entire U.S. population (approx. 327 million people), we would need to test 6.4 million people to have a reasonable assurance of detecting where the most infection cases are present.

Since it takes more than one test kit in some cases to test one person, we could be looking at 10 million test kits or more.

Not going to happen.

Even if we select a lower testing standard (i.e., a greater tolerance for missing some of the lesser infected areas), we’d still be looking at almost 5 million test kits.

A country that produces 1 billion cans of wet dog food every year can certainly muster up the resources to produce 5 million coronavirus test kits on short notice.

Furthermore, even if this country could produce them quickly, the resources necessary to distribute and administer these test kits would overwhelm our already fragile health care system.

Thus, this is why social distancing and self-quarantines are indispensable in containing the coronavirus and all future pathogens that will threaten us.

As Dr. Redlener noted on WNYC, we can’t detect and treat everyone.

_________________________________________________________________

Still, sampling methods such as acceptance sampling need to become standard practice among our front line health professionals, particularly at the beginning of a potential pandemic when its final social costs are still unknown.

The U.S. cannot afford to passively identify infected persons as we mostly have so far with COVID-19. Instead, the U.S. must proactively identify infected persons as South Korea and China are systematically doing now. We need to flatten the infection curve as best we can and sampling can help us efficiently do that.

Unfortunately, during the 2020 coronavirus pandemic, our state and federal governments appear unequipped and untrained to do that level of proactive detection.

As South Korea has proven, measurement is key to fighting pandemics like COVID-19.

Bob Galvin, a former Motorola executive and co-creator of the Lean Six Sigma quality improvement method, once said, “Companies must measure or die.” In the case of the COVID-19 pandemic, that warning is literally true.

• K.R.K.

Data sets and statistical codes referenced in this essay are available upon request to: kroeger98@yahoo.com

By Kent R. Kroeger (NuQum.com, March 14, 2020)

The importance of the vice presidential nominee has never been as great as it will be in the 2020 presidential election. The likely Democratic Party nominee — Joe Biden — is 77 years old and the probability his VP pick could be president before the next presidential election is (to put it gently) non-zero.

As for the incumbent, Donald Trump, the current consensus is that Vice President Mike Pence will be his running mate again in 2020. But if there is a time when an incumbent president should consider changing vice presidents, this is that time.

Pence offers no home state advantage. Trump would win Indiana if his running mate was a ham sandwich. Pence’s executive experience as a former Indiana governor is no longer necessary to compensate for Trump’s inexperience in a high-elected office. And don’t try to convince me that Pence is a charismatic campaigner who generates excitement within the Republican electoral base — he is duller than a wet paper towel.

But the real reason Trump should dump Pence has more to do with Trump’s electoral vulnerability than anything wrong with Pence. First, the coronavirus pandemic is starting to wear down Trump’s job approval — it now stands at 44.6 percent in the RealClearPolitics polling average, down in just a few weeks from 46 percent (the minimum level of approval Trump needs to have a decent chance at re-election, according to my recent analysis) — and without a significant shock in Trump’s favor before November, his chance of re-election is under grave threat.

Second, let us assume Biden will pick a woman as his running mate — according to one prediction market, PredictIt.com, she will be either Kamala Harris (29% chance), Amy Klobuchar (23% chance) or Stacey Abrams (15% chance). The media narrative coming out of the party conventions invariably will include the historic importance of the Biden’s VP selection and its appeal to a lot of swing voters, particularly women.

Trump’s argumentum ad captandum (‘appeal to the crowd’) would be significantly enhanced with his own trailblazing VP choice — which, I believe, will need to be either a woman, an African-American, or an Hispanic (or, even better, some combination of those identities).

The possible VP choices for Trump are fairly obvious (listed here in order of their odds based on my best hunch):

• Former US Ambassador to the UN Nikki Haley — my bet on Trump’s pick
• U.S. Senator Martha McSally (AZ)
• U.S. Representative Elise Stefanik (NY) — my personal favorite
• U.S. Senator Joni Ernst (IA)
• U.S. Senator Ted Cruz (TX)
• U.S. Senator Marco Rubio (FL)
• U.S. Senator Tim Scott (SC)

McSally, Ernst, Cruz and Rubio also happen to come from battleground states (though, technically, Texas is not one…yet) — which should enhance their attractiveness to Trump (more on that down below).

Stefanik, though from a noncompetitive blue state, might be the most intriguing person on the list. She’s young (age 35), telegenic, an energetic campaigner, and in possession of a growing national profile since her prominent role in the House impeachment hearings last fall.

I am far from the only person betting (or hoping) Pence is not going to see a second VP term. According to PredictIt.com, Pence’s probability of being re-nominated has been plummeting since he was selected to lead the Trump administration’s task force on the coronavirus — though he has regained a some support since Trump declared a national state of emergency (see chart below):

If Trump ever wanted to set up his Vice President to fail, making responsible for trying to stop the coronavirus would be at the top of the list. On the other hand, if this country were to somehow avoid the worst consequences of this pandemic (and that looks increasingly unlikely), Pence would not only enhance his value as Trump’s VP in the 2020 election, he’d be set up as the clear favorite for the 2024 Republican nomination.

________________________________________________________________

This discussion about presidential running mates may all be for naught.

Political scientists and election forecasters have long contented that, in general, vice presidential picks don’t affect U.S. presidential election outcomes.

“Few myths of American politics remain as deeply entrenched and blindly accepted as that of the electoral power of the vice presidential pick,” political scientists Kyle Kopko and Christopher Devine wrote in Politico in 2016.

In their research on whether a presidential candidate is more likely to win their running mate’s home-state, they found that no evidence to support such a conclusion: “Over the course of a little more than the past century, a vice presidential candidate’s state of residence generally has no effect on how a presidential candidate performs in that state. The vice presidential home state advantage is, essentially, zero.”

The same researchers found that VP choices don’t help much with specific demographic groups either.

Further evidence against the significance of VPs is the fact that none of the major econometric models used to forecast presidential elections include a variable for vice presidential picks; yet, the models still successfully forecast presidential election outcomes (most of the time). When the models do make mistakes, it is not because they failed to account for the VP selections.

Nonetheless, the belief in the importance of a VP pick continues in today’s election reporting, if only because a vice president can rise to the highest office due to the death or resignation of a president.

Vice presidential picks can furthermore help in other ways beyond a possible home state advantage. They can represent a critical constituency (Mike Pence in 2000 with evangelical voters), a ground-breaking candidacy (Geraldine Ferraro in 1984, Joe Lieberman in 2000), an augmentation to the qualifications and experience of the presidential candidate (Dick Cheney in 2000, Mike Pence 2016, Joe Biden 2008), or act as the campaign’s attack dog so the top of the ticket doesn’t need to descend into the mud (Richard Nixon 1952).

The political research against notwithstanding, in reality, there is concrete evidence that a VP selection can matter, and the case I cite most often is the 2000 presidential race between Al Gore and George W. Bush.

Most of us remember the 2000 election for how close it was and the ‘hanging chads’ in Florida. Few remember the vice presidential debate between Joe Lieberman and Dick Cheney that took place on October 5, 2000 at Centre College in Danville, Kentucky.

The hot take in the news media immediately after the VP debate was that it was civil, highly substantive, and probably boring to the average voter — hence, wouldn’t affect the final election outcome. The VP debate merely highlighted and amplified differences between Bush and Gore already known within the electorate, pundits said.

But something interesting and profound happened about 24 hours after the VP debate — tracking polls began to show a softening in Gore’s support (and a strengthening in Bush’s). The largest political futures market at the time — The Iowa Electronic Markets — showed an enduring, almost monotonic decline in the probability of Gore winning the election after the VP debate (see graph below).

What was the cause? There was no significant news event in that first week of October 2000 except for the VP debate.

Focus groups I conducted at the time while a graduate student at Columbia University offered an intriguing insight: Undecided voters found Lieberman’s raspy, Elmer Fudd-like voice unpleasant and distracting, while Cheney’s demeanor was perceived as comforting.

“He’s like a wise grandfather,” one female focus group participant said of Cheney. “I want to curl up in his lap and have him read me a story.”

That is probably not the image most of us have of Cheney today (if you do, I would keep it a secret); but, in the 2000 election, the Republicans had nominated a once-hard-drinking frat boy, George W. Bush, and many swing voters were uncomfortable with his anti-intellectual tendencies. Cheney wrapped these voters in a security blanket, giving them assurance that a junior Bush presidency would be run by adults, not just Delta Kappa Epsilon frat bothers.

The rest is history. You can decide if it was for better or worse.

My point is simply that VP choices can a have deep impact on the presidential vote. No research has convinced me otherwise.

If Donald Trump is as savvy as he tells us he is, he will understand how critical it is for his re-election chances to consider picking a new running mate.

Sorry, Mike. No time to worry about getting dumped from the GOP ticket. You have a far more important challenge facing you right now.

• K.R.K.

Comments and hot takes can be sent to: kroeger98@yahoo.com

By Kent R. Kroeger (NuQum.com, March 12, 2020)

Wealth, tourism and inexperience with communicable diseases are significant correlates with the current nation-level distribution of confirmed cases of COVID-19 (the disease caused by the coronavirus).

Using nation-level COVID-19 data from the World Health Organization (WHO) and socio-economic data from the World Bank, I developed a cross-sectional linear model to explain why, as of 10 March, some countries are seeing more confirmed cases of COVID-19 than others.

My initial findings are, on the one hand, unsurprising:

• All else equal, countries with larger populations have more confirmed cases of COVID-19;
• Countries with a high percentage of annual deaths from communicable diseases are seeing relatively fewer COVID-19 cases (mostly African, SE Asian, and Latin American countries);
• China has a disproportionate number of COVID-19 cases, even given its large population size;

And, in other cases, more thought-provoking:

• Countries with higher national incomes per capita are experiencing higher numbers of COVID-19 cases;
• Countries with higher numbers of tourists entering the country are experiencing higher numbers of COVID-19 cases.

[The linear regression model results can be found in the Appendix below]

The linear model accounts for 71 percent of the variance in confirmed COVID-19 cases on a national level.

Residual Analysis

Perhaps more interesting than those factors found to be significant correlates with COVID-19 cases were those countries where the model did not fit the data very well (see chart below).

Two countries in particular stand out in this chart: Turkey and the United States. Both have predicted values for the number of COVID-19 cases significantly higher than their reported numbers. As of 10 March, the U.S. reported 472 cases; whereas, the model suggests — based on the previously discussed variables — that the U.S. should have around 8,350 confirmed cases. Likewise, Turkey had not reported any cases as of 10 March, though the model predicts Turkey should have around 75 cases by now.

Given Turkey’s population size, economic wealth and sizable tourism industry, it is hard to believe there are no COVID-19 cases in the country. My presumption is that they are not testing for the coronavirus in any substantive way.

In the case of the U.S. we are faced with choosing one of two explanations: (1) Either the U.S. has done an unparalleled job of keeping the coronavirus from spreading, or (2) the U.S. has failed to adequately test for the coronavirus and thousands of Americans (possibly as many as 8,350 as of 10 March) have contracted this virus but have not been identified by the health care community and the CDC.

I want to believe the former explanation, but I’m afraid the answer is the latter.

With Vice President Mike Pence’s announcement last week that 1.1 million coronavirus tests would be distributed to qualified labs throughout the country, some optimism that the U.S. government is catching up to the gravity of the crisis. Unfortunately, the estimated U.S. infection numbers presented here tell a different story (see Table 1 below).

Using coronavirus test result numbers collected by British journalist Charlotte Gracias, the 1.1 million tests ordered by the U.S. government may not be near enough to handle the scale of this virus outbreak. If, in fact, there were 8,350 Americans with the coronavirus on 10 March, random testing at a national level would require over 500,000 Americans would need to be tested to find the positive cases. However, most subjects require two tests to determine if they carry the virus, meaning the U.S. will need over 1 million test kits to handle the magnitude of the outbreak on 10 March.

That was three days ago. Since then, the number of confirmed cases in the U.S. has risen from 472 to 1,268 (an over 160% increase). The U.S. government’s 1.1 million test kits won’t be completely distributed until next week (at the earliest). Yet, to handle this crisis, the U.S. already needs twice that number to handle the virus’ spread since VP Pence’s announcement.

This is a public health crisis unlike anything we’ve experienced since the 1918 Spanish Flu.

Implications

As this health crisis unfolds, I continue to test additional risk factor models using measures for volume of international trade, hospital beds/physicians per capita, quality of health care system, availability of affordable health care, demographic characteristics, income inequality, average climate, as well other variables. But so far, in every model I’ve tested, national wealth (per capita), tourism levels, and a nation’s previous experience with communicable diseases consistently maintained their statistical significance.

Of course, COVID-19 is an evolving crisis. These numbers will change over time and many of the countries not currently experiencing high numbers of COVID-19 cases (such as African and South American countries) may see their numbers rise dramatically — thereby, changing the model estimates.

In the meantime, it does appear national levels of wealth, tourism and inexperience with communicable diseases are significant upward drivers in the spread of COVID-19 worldwide.

Could it be that economic neoliberalism and globalization render us increasingly prone to these types of pandemics? Will viruses like the coronavirus become one of the recurring costs associated with the current international economic order?

The most likely answer is Yes.

_________________________________________________________________

Personal Thoughts on COVID-19

I once thought epidemics and pandemics were generally restricted to African countries and places where people eat monkeys and bats.

“I’ll just get a flu shot,” I would say to myself.

Its a callousness built by time and experience.

I lived through the Hong Kong flu of 1968–69, the Swine Flu of 1976, and the H1N1 Flu of 2009 — which is to say, I have no real memory of those epidemics. None of them impacted me personally. I vaguely recall getting in line at a nearby elementary school to get the Swine Flu vaccine in 1976 and being mad at my parents because I was missing an episode of Welcome Back, Kotter.

First world problems.

That has all changed with COVID-19 (which is caused by the virus 2019-nCoV, otherwise known as the coronavirus).

As of 12 March, there are 127,749 confirmed cases of COVID-19, resulting in 4,717 deaths worldwide. Those numbers will go up a lot before this pandemic ends.

According to infectious disease expert Dr. Michael Osterholm, director of the Center for Infectious Disease Research and Policy (CIDRAP) at the University of Minnesota, told podcaster Joe Rogan that COVID-19 could be “at least 10 to 15 times worse than the worst seasonal flu (in terms of fatalities).”

“We can conservatively estimate that this could require 48 million hospitalizations, 96 million cases actually occurring, and over 480 thousand deaths (worldwide),” says Dr. Osterholm.

A conservative estimate.

For me, the overwhelming emotion right now is that of powerlessness. Beyond the obvious but impractical response — not leaving the house until this is over — I don’t know what else to do to protect my family and myself. It feels like a game of Russian roulette on a global scale. We could do everything possible — wear a mask and gloves, avoid crowds, etc. — and still get the virus.

As a statistician, I bring a clinical, dispassionate understanding of random chance and relative risk to this crisis. As a husband and parent, I am staggered by a slow-burning anxiety these days.

I just want this to be over.

• K.R.K.

Comments and suggestions can be sent to: kroeger98@yahoo.com

_________________________________________________________________

Appendix: A Regression Model of Confirmed COVID-19 Cases Worldwide (as of 10 March 2020)

Variables

Dependent Variable (Source: WHO): Number of confirmed COVID-19 cases as of 10 March 2020 (log transformed)

Independent Variables (Source: World Bank):

NY_ADJ_NNTY_PC_CD = Adjusted net nat’l income per capita (current US$)

LOG_POP = Country’s total populaton (log transformed)

CHINA = Indicator variable (1 = China; 0 = all other countries)

SI_INT_ARVL = International tourism, number of arrivals

SH_DTH_COMM_ZS = Cause of death, by communicable diseases and maternal, prenatal and nutrition conditions (% of all deaths)

More detail on these independent variables can be found in the World Bank’s data catalog at: https://datacatalog.worldbank.org/dataset/world-development-indicators

Model Fit

Overall, the model explained 71 percent of the variance in confirmed COVID-19 cases on a national level.

Other Methodological Issues:

I should note that I used a linear regression model in this analysis. However, a Poisson model is the appropriate technique to use with count-data (such as the number of COVID-19 cases). I chose to report the linear regression model, in part, because linear regression is generally more familiar to a larger segment of readers. Nonetheless, I did run the same model using Poisson regression in SPSS and found no substantive differences in the results.